Lang Gao, Xuhui Li, Chenxi Wang, Mingzhe Li, Wei Liu, Zirui Song, Jinghui Zhang, Rui Yan, Preslav Nakov, Xiuying Chen. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2026.

The recent advancement in video temporal grounding (VTG) has significantly enhanced fine-grained video understanding, primarily driven by multimodal large language models (MLLMs). With superior multimodal comprehension and reasoning abilities, VTG approaches based on MLLMs (VTG-MLLMs) are gradually surpassing traditional fine-tuned methods. They not only achieve competitive performance but also excel in generalization across zero-shot, multi-task, and multi-domain settings. Despite extensive surveys on general video-language understanding, comprehensive reviews specifically addressing VTG-MLLMs remain scarce. To fill this gap, this survey systematically examines current research on VTG-MLLMs through a three-dimensional taxonomy: 1) the functional roles of MLLMs, highlighting their architectural significance; 2) training paradigms, analyzing strategies for temporal reasoning and task adaptation; and 3) video feature processing techniques, which determine spatiotemporal representation effectiveness. We further discuss benchmark datasets, evaluation protocols, and summarize empirical findings. Finally, we identify existing limitations and propose promising research directions.

Image synthesis is a key application of generative AI. It can help reduce overfitting and the high cost of collecting real-world data for downstream discriminative models. However, current methods mainly focus on making images look realistic and ignore their true goal: improving downstream model generalization and robustness. We find that existing approaches tend to produce high-fidelity synthetic images that closely resemble the original data. This limits their value for improving downstream task performance. To overcome this, we argue that diversity, not just fidelity, must guide synthetic data generation if it is to truly complement human-collected datasets. In this paper, we introduce a Retrieval-Augmented Generation framework for diverse diffusion-based image synthesis. At each generation step, we retrieve the top-K most similar samples in feature space from both real and previously generated images. We then apply an Anti-Attention mechanism that actively pushes the new image away from these retrieved samples in feature space, maximizing dissimilarity. We propose novel evaluation metrics to assess image synthesis diversity and demonstrate significant improvements over existing benchmarks. Moreover, downstream models trained with our synthetic data achieved a 1.9% absolute accuracy gain on standard benchmarks, outperforming existing synthesis techniques.

Whole-body multimodal motion generation, controlled by text, speech, or music, has numerous applications including video generation and character animation. However, employing a unified model to process different condition modalities presents two main challenges: motion distribution drifts across different tasks (e.g., co-speech gestures and text-driven daily actions) and the complex optimization of mixed conditions with varying granularities (e.g., text and audio). In this paper, we propose MotionCraft, a unified diffusion transformer that crafts whole-body motion with plug-and-play multimodal control. Our framework employs a coarse-to-fine training strategy, starting with the text-to-motion semantic pre-training, followed by the multimodal low-level control adaptation. To effectively learn and transfer motion knowledge across different distributions, we design MC-Attn for parallel modeling of static and dynamic human topology graphs. To overcome the motion format inconsistency of existing benchmarks, we introduce MC-Bench, the first available multimodal whole-body motion generation benchmark based on the unified SMPL-X format. Extensive experiments show that MotionCraft achieves state-of-the-art performance on various standard motion generation tasks.

This paper explores higher-resolution video outpainting with extensive content generation. We point out common issues faced by existing methods when attempting to largely outpaint videos: the generation of low-quality content and limitations imposed by GPU memory. To address these challenges, we propose a diffusion-based method called Infinite-Canvas. It builds upon two core designs. First, instead of employing the common practice of "single-shot" outpainting, we distribute the task across spatial windows and seamlessly merge them. It allows us to outpaint videos of any size and resolution without being constrained by GPU memory. Second, the source video and its relative positional relation are injected into the generation process of each window. It makes the generated spatial layout within each window harmonize with the source video. Coupling with these two designs enables us to generate higher-resolution outpainting videos with rich content while keeping spatial and temporal consistency. Infinite-Canvas excels in large-scale video outpainting, e.g., from 512 × 512 to 1152 × 2048 (9×), while producing high-quality and aesthetically pleasing results. It achieves the best quantitative results across various resolution and scale setups. The code is available at https://github.com/mayuelala/FollowYourCanvas.

Bayesian Optimization (BO) is a sample-efficient black-box optimizer commonly used in search spaces where hyperparameters are independent. However, in many practical AutoML scenarios, there will be dependencies among hyperparameters, forming a conditional search space, which can be partitioned into structurally distinct subspaces. The structure and dimensionality of hyperparameter configurations vary across these subspaces, challenging the application of BO. Some previous BO works have proposed solutions to develop multiple Gaussian Process models in these subspaces. However, these approaches tend to be inefficient as they require a substantial number of observations to guarantee each GP's performance and cannot capture relationships between hyperparameters across different subspaces. To address these issues, this paper proposes a novel approach to model the response surfaces of all subspaces in one, which can model the relationships between hyperparameters elegantly via a self-attention mechanism. Concretely, we design a structure-aware hyperparameter embedding to preserve the structural information. Then, we introduce an attention-based deep feature extractor, capable of projecting configurations with different structures from various subspaces into a unified feature space, where the response surfaces can be formulated using a single standard Gaussian Process. The empirical results on a simulation function, various real-world tasks, and HPO-B benchmark demonstrate that our proposed approach improves the efficacy and efficiency of BO within conditional search spaces.

Despite recent advances in image-to-video generation, better controllability and local animation are less explored. Most existing image-to-video methods are not locally aware and tend to move the entire scene. However, human artists may need to control the movement of different objects or regions. Additionally, current I2V methods require users not only to describe the target motion but also to provide redundant detailed descriptions of frame contents.These two issues hinder the practical utilization of current I2V tools. In this paper, we propose a practical framework, named Follow-Your-Click, to achieve image animation with a simple user click (for specifying what to move) and a motion prompt (for specifying how to move). Technically, we propose the first-frame masking strategy, which significantly improves the video generation quality, and a motion-augmented module equipped with a motion prompt dataset to improve the motion prompt following abilities of our model. To further control the motion speed, we propose flow-based motion magnitude control to control the speed of target movement more precisely. Extensive experiments compared with 7 baselines, including both commercial tools and research methods on 8 metrics, suggest the superiority of our approach.

Diffusion models have exhibited impressive prowess in the text-to-image task. Recent methods add image-level structure controls, e.g., edge and depth maps, to manipulate the generation process together with text prompts to obtain desired images. This controlling process is globally operated on the entire image, which limits the flexibility of control regions. In this paper, we explore a novel and practical task setting: local control. It focuses on controlling specific local region according to user-defined image conditions, while the remaining regions are only conditioned by the original text prompt. However, it is non-trivial to achieve it. The naive manner of directly adding local conditions may lead to the local control dominance problem, which forces the model to focus on the controlled region and neglect object generation in other regions. To mitigate this problem, we propose Regional Discriminate Loss to update the noised latents, aiming at enhanced object generation in non-control regions. Furthermore, the proposed Focused Token Response suppresses weaker attention scores which lack the strongest response to enhance object distinction and reduce duplication. Lastly, we adopt Feature Mask Constraint to reduce quality degradation in images caused by information differences across the local control region. All proposed strategies are operated at the inference stage. Extensive experiments demonstrate that our method can synthesize high-quality images aligned with the text prompt under local control conditions.

Peng Zhou, Pengsen Ma, Jianmin Wang, Xibao Cai, Haitao Huang, Wei Liu, Longyue Wang, Lai Hou Tim, Xiangxiang Zeng. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2025.

Creating a high-fidelity, animatable 3D full-body avatar from a single image is a challenging task due to the diverse appearance and poses of humans and the limited availability of high-quality training data. To achieve fast and high-quality human reconstruction, this work rethinks the task from the perspectives of dataset, model, and representation. First, we introduce a large-scale HUman-centric GEnerated dataset, HuGe100K, consisting of 100K diverse, photorealistic sets of human images. Each set contains 24-view frames in specific human poses, generated using a pose-controllable image-to-multi-view model. Next, leveraging the diversity in views, poses, and appearances within HuGe100K, we develop a scalable feed-forward transformer model to predict a 3D human Gaussian representation in a uniform space from a given human image. This model is trained to disentangle human pose, body shape, clothing geometry, and texture. The estimated Gaussians can be animated without post-processing. We conduct comprehensive experiments to validate the effectiveness of the proposed dataset and method. Our model demonstrates the ability to efficiently reconstruct photorealistic humans at 1K resolution from a single input image using a single GPU instantly. Additionally, it seamlessly supports various applications, as well as shape and texture editing tasks.

Controllable character image animation has a wide range of applications. Although existing studies have consistently improved performance, challenges persist in the field of character image animation, particularly concerning stability in complex backgrounds and tasks involving multiple characters. To address these challenges, we propose a novel multi-condition guided framework for character image animation, employing several well-designed input modules to enhance the implicit decoupling capability of the model. First, the optical flow guider calculates the background optical flow map as guidance information, which enables the model to implicitly learn to decouple the background motion into background constants and background momentum during training, and generate a stable background by setting zero background momentum during inference. Second, the depth order guider calculates the order map of the characters, which transforms the depth information into the positional information of multiple characters. This facilitates the implicit learning of decoupling different characters, especially in accurately separating the occluded body parts of multiple characters. Third, the reference pose map is input to enhance the ability to decouple character texture and pose information in the reference image. Furthermore, to fill the gap of fair evaluation of multi-character image animation, we propose a new benchmark comprising about 4,000 frames. Extensive qualitative and quantitative evaluations demonstrate that our method excels in generating high-quality character animations, especially in scenarios of complex backgrounds and multiple characters.

Molecule-and-text cross-modal representation learning has emerged as a promising direction for enhancing the quality of molecular representation, thereby improving performance in various scientific fields. However, most approaches employ a global alignment approach to learn the knowledge from different modalities that may fail to capture fine-grained information, such as molecule-and-text fragments and stereoisomeric nuances, which is crucial for downstream tasks. Furthermore, it is incapable of modeling such information using a similar global alignment strategy due to the lack of annotations about the fine-grained fragments in the existing dataset. In this paper, we propose Atomas, a hierarchical molecular representation learning framework that jointly learns representations from SMILES strings and text. We design a Hierarchical Adaptive Alignment model to automatically learn the fine-grained fragment correspondence between two modalities and align these representations at three semantic levels. Atomas's end-to-end training framework supports understanding and generating molecules, enabling a wider range of downstream tasks. Atomas achieves superior performance across 12 tasks on 11 datasets, outperforming 11 baseline models thus highlighting the effectiveness and versatility of our method. Scaling experiments further demonstrate Atomas’s robustness and scalability. Moreover, visualization and qualitative analysis, validated by human experts, confirm the chemical relevance of our approach. Codes are released on ~\url{https://github.com/yikunpku/Atomas}.

Although significant advancements have been achieved in the progress of keypoint-guided Text-to-Image diffusion models, existing mainstream keypoint-guided models encounter challenges in controlling the generation of more general non-rigid objects beyond humans (e.g., animals). Moreover, it is difficult to generate multiple overlapping humans and animals based on keypoint controls solely. These challenges arise from two main aspects: the inherent limitations of existing controllable methods and the lack of suitable datasets. First, we design a DiT-based framework, named UniMC, to explore unifying controllable multi-class image generation. UniMC integrates instance- and keypoint-level conditions into compact tokens, incorporating attributes such as class, bounding box, and keypoint coordinates. This approach overcomes the limitations of previous methods that struggled to distinguish instances and classes due to their reliance on skeleton images as conditions. Second, we propose HAIG-2.9M, a large-scale, high-quality, and diverse dataset designed for keypoint-guided human and animal image generation. HAIG-2.9M includes 786K images with 2.9M instances. This dataset features extensive annotations such as keypoints, bounding boxes, and fine-grained captions for both humans and animals, along with rigorous manual inspection to ensure annotation accuracy. Extensive experiments demonstrate the high quality of HAIG-2.9M and the effectiveness of UniMC, particularly in heavy occlusions and multi-class scenarios.

Recently, the impressive performance of large language models (LLMs) on a wide range of tasks has attracted an increasing number of attempts to apply LLMs in drug discovery. However, molecule optimization, a critical task in the drug discovery pipeline, is currently an area that has seen little involvement from LLMs. Most of existing approaches focus solely on capturing the underlying patterns in chemical structures provided by the data, without taking advantage of expert feedback. These non-interactive approaches overlook the fact that the drug discovery process is actually one that requires the integration of expert experience and iterative refinement. To address this gap, we propose DrugAssist, an interactive molecule optimization model which performs optimization through human-machine dialogue by leveraging LLM's strong interactivity and generalizability. DrugAssist has achieved leading results in both single and multiple property optimization, simultaneously showcasing immense potential in transferability and iterative optimization. In addition, we publicly release a large instruction-based dataset called 'MolOpt-Instructions' for fine-tuning language models on molecule optimization tasks. We have made our code and data publicly available at https://github.com/blazerye/DrugAssist, which we hope to pave the way for future research in LLMs' application for drug discovery.

Masked autoencoders (MAEs) have recently achieved great success in computer vision. They can automatically extract representations from unlabeled data and improve the performance of various downstream tasks. However, training an MAE model requires substantial resources, which limits their accessibility to many academic institutions: often laboratories in universities lack the necessary resources. This issue significantly hinders the development of this field. In this paper, we propose FastMAE, an efficient MAE approach. Inspired by the idea of offline tokenizers in natural language processing, FastMAE presents a novel way to build an offline vision tokenizer, which can provide high-level semantics in an efficient way. Benefiting from the offline tokenizer, FastMAE becomes an efficient vision learner. Our experiments demonstrate that FastMAE can achieve 83.6% accuracy with ViT-B in only 18.8 h on 8 NVIDIA Tesla-V100 GPUs, which is 31.3× faster than the original MAE, providing a resource friendly baseline for the computer vision community. Moreover, it also achieves comparable performance to state-of-the-art methods. We hope our research will attract more people to engage in MAE-related research and that we can advance its development together.

Assessing risk factors for intracranial aneurysm (IA) development on images is crucial for early detection of high-risk cases. IAs often form at bifurcations within the circle of Willis (CoW), but manual assessment of these arteries is both time-consuming and susceptible to inconsistencies. Previous studies on imaging markers for IA development lack sufficient evidence for clinical implications, highlighting the need for automated methods to assess CoW morphology. No systematic approach currently exists to identify the best methodological strategies. To address this, we organized a scientific challenge to compare various techniques against a clinical reference standard. Participants were tasked with (1) automated classification of CoW anatomical variants and (2) automated prediction of CoW artery diameters and bifurcation angles. We provided 300 TOF-MRA scans for training and another 300 for testing, all manually annotated. Submissions were evaluated using balanced accuracy, mean absolute error, and Pearson correlation coefficient metrics. This paper provides a detailed analysis of the results from six participating teams. The findings show that various methods may be suitable for automated CoW assessment, but that these need further improvement to meet clinical standards. The challenge remains open for future submissions, offering a benchmark for new techniques.

Cross-modal alignment plays a crucial role in vision-language pre-training (VLP) models, enabling them to capture meaningful associations across different modalities. For this purpose, inspired by the success of masked language modeling (MLM) tasks in the NLP pre-training area, numerous masked modeling tasks have been proposed for VLP to further promote cross-modal interactions. The core idea of previous masked modeling tasks is to focus on reconstructing the masked tokens based on visible context for learning local-local alignment, i.e., associations between image patches and text tokens. However, most of them pay little attention to the global semantic features generated for the masked data, resulting in a limited cross-modal alignment ability of global representations to local features of the other modality. Therefore, in this paper, we propose a novel Global and Local Semantic Completion Learning (GLSCL) task to facilitate global-local alignment and local-local alignment simultaneously. Specifically, the GLSCL task complements the missing semantics of masked data and recovers global and local features by cross-modal interactions. Our GLSCL consists of masked global semantic completion (MGSC) and masked local token completion (MLTC). MGSC promotes learning more representative global features, which have a great impact on the performance of downstream tasks, while MLTC reconstructs modal-fusion local tokens, further enhancing accurate comprehension of multimodal data. To evaluate the proposed approaches on cross-modal alignment, we develop a validation benchmark called ALIGN-BENCH. Moreover, we present a flexible vision encoder, enabling our model to simultaneously perform image-text and video-text multimodal tasks. Experimental results show that our proposed method obtains state-of-the-art performance on various vision-language benchmarks, such as visual question answering, image-text retrieval, and video-text retrieval.

Customization generation techniques have significantly advanced the synthesis of specific concepts across varied contexts. Multi-concept customization emerges as the challenging task within this domain. Existing approaches often rely on training a fusion matrix of multiple Low-Rank Adaptations (LoRAs) to merge various concepts into a single image. However, we identify this straightforward method faces two major challenges: 1) concept confusion, where the model struggles to preserve distinct individual characteristics, and 2) concept vanishing, where the model fails to generate the intended subjects. To address these issues, we introduce LoRA-Composer, a training-free framework designed for seamlessly integrating multiple LoRAs, thereby enhancing the harmony among different concepts within generated images. LoRA-Composer addresses concept vanishing through concept injection constraints, enhancing visibility via an expanded cross-attention mechanism. To combat concept confusion, concept isolation constraints are introduced, refining the self-attention computation. Furthermore, we propose two inference techniques to accelerate inference speed without performance degradation and enhance the accuracy of the generated region, respectively. Extensive experiments demonstrate that LoRA-Composer significantly outperforms standard baselines, especially in scenarios without image-based conditions such as canny edge or pose estimation.

Contrastive Vision-Language Pre-training, known as CLIP, has shown promising effectiveness in addressing downstream image recognition tasks. However, recent works revealed that the CLIP model can be implanted with a downstream-oriented backdoor. On downstream tasks, one victim model performs well on clean samples but predicts a specific target class whenever a specific trigger is present. For injecting a backdoor, existing attacks depend on a large amount of additional data to maliciously fine-tune the entire pre-trained CLIP model, which makes them inapplicable to data-limited scenarios. In this work, motivated by the recent success of learnable prompts, we address this problem by injecting a backdoor into the CLIP model in the prompt learning stage. Our method named BadCLIP is built on a novel and effective mechanism in backdoor attacks on CLIP, i.e., influencing both the image and text encoders with the trigger. It consists of a learnable trigger applied to images and a trigger-aware context generator, such that the trigger can change text features via trigger-aware prompts, resulting in a powerful and generalizable attack. Extensive experiments conducted on 11 datasets verify that the clean accuracy of BadCLIP is similar to those of advanced prompt learning methods and the attack success rate is higher than 99% in most cases. BadCLIP is also generalizable to unseen classes, and shows a strong generalization capability under cross-dataset and cross-domain settings. The code is available at https://github.com/jiawangbai/BadCLIP.

Existing game AI research mainly focuses on enhancing agents' abilities to win games, but this does not inherently make humans have a better experience when collaborating with these agents. For example, agents may dominate the collaboration and exhibit unintended or detrimental behaviors, leading to poor experiences for their human partners. In other words, most game AI agents are modeled in a "self-centered" manner. In this paper, we propose a "human-centered" modeling scheme for collaborative agents that aims to enhance the experience of humans. Specifically, we model the experience of humans as the goals they expect to achieve during the task. We expect that agents should learn to enhance the extent to which humans achieve these goals while maintaining agents' original abilities (e.g., winning games). To achieve this, we propose the Reinforcement Learning from Human Gain (RLHG) approach. The RLHG approach introduces a "baseline", which corresponds to the extent to which humans primitively achieve their goals, and encourages agents to learn behaviors that can effectively enhance humans in achieving their goals better. We evaluate the RLHG agent in the popular Multi-player Online Battle Arena (MOBA) game, Honor of Kings, by conducting real-world human-agent tests. Both objective performance and subjective preference results show that the RLHG agent provides participants better gaming experience.

Large vision-language models (VLMs) such as GPT-4 have achieved exceptional performance across various multi-modal tasks. However, the deployment of VLMs necessitates substantial energy consumption and computational resources. Once attackers maliciously induce high energy consumption and latency time (energy-latency cost) during inference of VLMs, it will exhaust computational resources. In this paper, we explore this attack surface about availability of VLMs and aim to induce high energy-latency cost during inference of VLMs. We find that high energy-latency cost during inference of VLMs can be manipulated by maximizing the length of generated sequences. To this end, we propose verbose images, with the goal of crafting an imperceptible perturbation to induce VLMs to generate long sentences during inference. Concretely, we design three loss objectives. First, a loss is proposed to delay the occurrence of end-of-sequence (EOS) token, where EOS token is a signal for VLMs to stop generating further tokens. Moreover, an uncertainty loss and a token diversity loss are proposed to increase the uncertainty over each generated token and the diversity among all tokens of the whole generated sequence, respectively, which can break output dependency at token-level and sequence-level. Furthermore, a temporal weight adjustment algorithm is proposed, which can effectively balance these losses. Extensive experiments demonstrate that our verbose images can increase the length of generated sequences by 7.87× and 8.56× compared to original images on MS-COCO and ImageNet datasets, which presents potential challenges for various applications.

Protein complex formation, a pivotal challenge in contemporary biology, has recently gained interest from the machine learning community, particularly concerning protein-ligand docking tasks. In this paper, we delve into the equally crucial but comparatively under-investigated domain of protein-protein docking. Specifically, we propose a geometric deep learning framework, termed EBMDock, which employs statistical potential as its energy function. This approach produces a probability distribution over docking poses, such that the identified docking pose aligns with a minimum point in the energy landscape. We employ a differential algorithm grounded in Langevin dynamics to efficiently sample from the docking pose distribution. Additionally, we incorporate energy-based training using contrastive divergence, enhancing both performance and stability. Empirical results demonstrate that our approach achieves superior performance on two benchmark datasets DIPS and DB5.5. Furthermore, the results suggest EBMDock can serve as an orthogonal enhancement to existing methods.

The video-language (VL) pretraining has achieved remarkable improvement in multiple downstream tasks. However, the current VL pretraining framework is hard to extend to multiple modalities (N modalities, N ≥ 3) beyond vision and language. We thus propose LanguageBind, taking the language as the bind across different modalities because the language modality is well-explored and contains rich semantics. Specifically, we freeze the language encoder acquired by VL pretraining and then train encoders for other modalities with contrastive learning. As a result, all modalities are mapped to a shared feature space, implementing multi-modal semantic alignment. While LanguageBind ensures that we can extend VL modalities to N modalities, we also need a high-quality dataset with alignment data pairs centered on language. We thus propose VIDAL-10M with 10 Million data with Video, Infrared, Depth, Audio and their corresponding Language. In our VIDAL-10M, all videos are from short video platforms with complete semantics rather than truncated segments from long videos, and all the video, depth, infrared, and audio modalities are aligned to their textual descriptions. LanguageBind has achieved superior performance on a wide range of 15 benchmarks covering video, audio, depth, and infrared. Moreover, multiple experiments have provided evidence for the effectiveness of LanguageBind in achieving indirect alignment and complementarity among diverse modalities.

Diffusion models have demonstrated their effectiveness in addressing the inherent uncertainty and indeterminacy in monocular 3D human pose estimation (HPE). Despite their strengths, the need for large search spaces and the corresponding demand for substantial training data make these models prone to generating biomechanically unrealistic poses. This challenge is particularly noticeable in occlusion scenarios, where the complexity of inferring 3D structures from 2D images intensifies. In response to these limitations, we introduce the **Di**screte **Di**ffusion **Pose** (**$\text{Di}^2\text{Pose}$**), a novel framework designed for occluded 3D HPE that capitalizes on the benefits of a discrete diffusion model. Specifically, **$\text{Di}^2\text{Pose}$** employs a two-stage process: it first converts 3D poses into a discrete representation through a pose quantization step, which is subsequently modeled in latent space through a discrete diffusion process. This methodological innovation restrictively confines the search space towards physically viable configurations and enhances the model’s capability to comprehend how occlusions affect human pose within the latent space. Extensive evaluations conducted on various benchmarks (e.g., Human3.6M, 3DPW, and 3DPW-Occ) have demonstrated its effectiveness.

We present Follow-Your-Emoji, a diffusion-based framework for portrait animation, which animates a reference portrait with target landmark sequences. The main challenge of portrait animation is to preserve the identity of the reference portrait and transfer the target expression to this portrait while maintaining temporal consistency and fidelity. To address these challenges, Follow-Your-Emoji equipped the powerful Stable Diffusion model with two well-designed technologies. Specifically, we first adopt a new explicit motion signal, namely expression-aware landmark, to guide the animation process. We discover this landmark can not only ensure the accurate motion alignment between the reference portrait and target motion during inference but also increase the ability to portray exaggerated expressions (i.e., large pupil movements) and avoid identity leakage. Then, we propose a facial fine-grained loss to improve the model’s ability of subtle expression perception and reference portrait appearance reconstruction by using both expression and facial masks. Accordingly, our method demonstrates significant performance in controlling the expression of freestyle portraits, including real humans, cartoons, sculptures, and even animals. By leveraging a simple and effective progressive generation strategy, we extend our model to stable long-term animation, thus increasing its potential application value. To address the lack of a benchmark for this field, we introduce EmojiBench, a comprehensive benchmark comprising diverse portrait images, driving videos, and landmarks. We show extensive evaluations on EmojiBench to verify the superiority of Follow-Your-Emoji. The code, training dataset and benchmark will be found in https://github.com/mayuelala/FollowYourEmoji.

Nearly all existing scene graph generation (SGG) models have overlooked the ground-truth annotation qualities of mainstream SGG datasets, i.e., they assume: 1) all the manually annotated positive samples are equally correct; 2) all the un-annotated negative samples are absolutely background. In this paper, we argue that neither of the assumptions applies to SGG: there are numerous “noisy” ground-truth predicate labels that break these two assumptions and harm the training of unbiased SGG models. To this end, we propose a novelNoIsy labelCorrEction andSampleTrainingstrategy for SGG:NICEST, which rules out these noisy label issues by generating high-quality samples and designing an effective training strategy. Specifically, it consists of: 1)NICE: it detects noisy samples and then reassigns higher-quality soft predicate labels to them. To achieve this goal, NICE contains three main steps: negative Noisy Sample Detection (Neg-NSD), positive NSD (Pos-NSD), and Noisy Sample Correction (NSC). Firstly, in Neg-NSD, it is treated as an out-of-distribution detection problem, and the pseudo labels are assigned to all detected noisy negative samples. Then, in Pos-NSD, we use a density-based clustering algorithm to detect noisy positive samples. Lastly, in NSC, we use weighted KNN to reassign more robust soft predicate labels rather than hard labels to all noisy positive samples. 2)NIST: it is a multi-teacher knowledge distillation based training strategy, which enables the model to learn unbiased fusion knowledge. A dynamic trade-off weighting strategy in NIST is designed to penalize the bias of different teachers. Due to the model-agnostic nature of both NICE and NIST, NICEST can be seamlessly incorporated into any SGG architecture to boost its performance on different predicate categories. In addition, to better assess the generalization ability of SGG models, we propose a new benchmark,VG-OOD, by reorganizing the prevalent VG dataset. This reorganization deliberately makes the predicate distributions between the training and test sets as different as possible for each subject-object category pair. This new benchmark helps disentangle the influence of subject-object category biases. Extensive ablations and results on different backbones and tasks have attested to the effectiveness and generalization ability of each component of NICEST.

Rank aggregation with pairwise comparisons is widely encountered in sociology, politics, economics, psychology, sports, etc. Given the enormous social impact and the consequent incentives, the potential adversary has a strong motivation to manipulate the ranking list. However, the ideal attack opportunity and the excessive adversarial capability cause the existing methods to be impractical. To fully explore the potential risks, we leverage an online attack on the vulnerable data collection process. Since it is independent of rank aggregation and lacks effective protection mechanisms, we disrupt the data collection process by fabricating pairwise comparisons without knowledge of the future data or the true distribution. From the game-theoretic perspective, the confrontation scenario between the online manipulator and the ranker who takes control of the original data source is formulated as a distributionally robust game that deals with the uncertainty of knowledge. Then we demonstrate that the equilibrium in the above game is potentially favorable to the adversary by analyzing the vulnerability of the sampling algorithms such as Bernoulli and reservoir methods. According to the above theoretical analysis, different sequential manipulation policies are proposed under a Bayesian decision framework and a large class of parametric pairwise comparison models. For attackers with complete knowledge, we establish the asymptotic optimality of the proposed policies. To increase the success rate of the sequential manipulation with incomplete knowledge, a distributionally robust estimator, which replaces the maximum likelihood estimation in a saddle point problem, provides a conservative data generation solution. Finally, the corroborating empirical evidence shows that the proposed method manipulates the results of rank aggregation methods in a sequential manner.

While features of different scales are perceptually important to visual inputs, existing vision transformers do not yet take advantage of them explicitly. To this end, we first propose a cross-scale vision transformer, CrossFormer. It introduces a cross-scale embedding layer (CEL) and a long-short distance attention (LSDA). On the one hand, CEL blends each token with multiple patches of different scales, providing the self-attention module itself with cross-scale features. On the other hand, LSDA splits the self-attention module into a short-distance one and a long-distance counterpart, which not only reduces the computational burden but also keeps both small-scale and large-scale features in the tokens. Moreover, through experiments on CrossFormer, we observe another two issues that affect vision transformers' performance, i.e., the enlarging self-attention maps and amplitude explosion. Thus, we further propose a progressive group size (PGS) paradigm and an amplitude cooling layer (ACL) to alleviate the two issues, respectively. The CrossFormer incorporating with PGS and ACL is called CrossFormer++. Extensive experiments show that CrossFormer++ outperforms the other vision transformers on image classification, object detection, instance segmentation, and semantic segmentation tasks.

Low-light image enhancement (LLIE) investigates how to improve the brightness of an image captured in illumination-insufficient environments. The majority of existing methods enhance low-light images in a global and uniform manner, without taking into account the semantic information of different regions. Consequently, a network may easily deviate from the original color of local regions. To address this issue, we propose a semantic-aware knowledge-guided framework (SKF) that can assist a low-light enhancement model in learning rich and diverse priors encapsulated in a semantic segmentation model. We concentrate on incorporating semantic knowledge from three key aspects: a semantic-aware embedding module that adaptively integrates semantic priors in feature representation space, a semantic-guided color histogram loss that preserves color consistency of various instances, and a semantic-guided adversarial loss that produces more natural textures by semantic priors. Our SKF is appealing in acting as a general framework in the LLIE task. We further present a refined framework SKF++ with two new techniques: (a) Extra convolutional branch for intra-class illumination and color recovery through extracting local information and (b) Equalization-based histogram transformation for contrast enhancement and high dynamic range adjustment. Extensive experiments on various benchmarks of LLIE task and other image processing tasks show that models equipped with the SKF/SKF++ significantly outperform the baselines and our SKF/SKF++ generalizes to different models and scenes well. Besides, the potential benefits of our method in face detection and semantic segmentation in low-light conditions are discussed.

In the real world, how to effectively learn consistent similarity measurement across different modalities is essential. Most of the existing similarity learning methods cannot deal well with cross-modal data due to the modality gap and have obvious performance degeneration when applied to cross-modal data. To tackle this problem, we propose a novel cross-modal similarity learning method, called Causality-Invariant Interactive Mining (CIIM), that can effectively capture informative relationships among different samples and modalities to derive the modality-consistent feature embeddings in the unified metric space. Our CIIM tackles the modality gap from two aspects, i.e., sample-wise and feature-wise. Specifically, we start from the sample-wise view and learn the single-modality and hybrid-modality proxies for exploring the cross-modal similarity with the elaborate metric losses. In this way, sample-to-sample and sample-to-proxy correlations are both taken into consideration. Furthermore, we conduct the causal intervention to eliminate the modality bias and reconstruct the invariant causal embedding in the feature-wise aspect. To this end, we force the learned embeddings to satisfy the specific properties of our causal mechanism and derive the causality-invariant feature embeddings in the unified metric space. Extensive experiments on two cross-modality tasks demonstrate the superiority of our proposed method over the state-of-the-art methods.

Mining discriminative graph topological information plays an important role in promoting graph representation ability. However, it suffers from two main issues: (1) the difficulty/complexity of computing global inter-class/intra-class scatters, commonly related to mean and covariance of graph samples, for discriminant learning; (2) the huge complexity and variety of graph topological structure that is rather challenging to robustly characterize. In this paper, we propose the Wasserstein Discriminant Dictionary Learning (WDDL) framework to achieve discriminant learning on graphs with robust graph topology modeling, and hence facilitate graph-based pattern analysis tasks. Considering the difficulty of calculating global inter-class/intra-class scatters, a reference set of graphs (aka graph dictionary) is first constructed by generating representative graph samples (aka graph keys) with expressive topological structure. Then, a Wasserstein Graph Representation (WGR) process is proposed to project input graphs into a succinct dictionary space through the graph dictionary lookup. To further achieve discriminant graph learning, a Wasserstein discriminant loss (WD-loss) is defined on the graph dictionary, in which the graph keys are optimizable, to make the intra-class keys more compact and inter-class keys more dispersed. Hence, the calculation of global Wasserstein metric (W-metric) centers can be bypassed. For sophisticated topology mining in the WGR process, a joint-Wasserstein graph embedding module is constructed to model both between-node and between-edge relationships across inputs and graph keys by encapsulating both the Wasserstein metric (between cross-graph nodes) and proposed novel Kron-Gromov-Wasserstein (KGW) metric (between cross-graph adjacencies). Specifically, the KGW-metric comprehensively characterizes the cross-graph connection patterns with the Kronecker operation, then adaptively captures those salient patterns through connection pooling. To evaluate the proposed framework, we study two graph-based pattern analysis problems, i.e. graph classification and cross-modal retrieval, with the graph dictionary flexibly adjusted to cater to these two tasks. Extensive experiments are conducted to comprehensively compare with existing advanced methods, as well as dissect the critical component of our proposed architecture. The experimental results validate the effectiveness of the WDDL framework.

Blur artifacts can seriously degrade the visual quality of images, and numerous deblurring methods have been proposed for specific scenarios. However, in most real-world images, blur is caused by different factors, e.g., motion, and defocus. In this paper, we address how other deblurring methods perform in the case of multiple types of blur. For in-depth performance evaluation, we construct a new large-scale multi-cause image deblurring dataset (MC-Blur), including real-world and synthesized blurry images with different blur factors. The images in the proposed MC-Blur dataset are collected using other techniques: averaging sharp images captured by a 1000-fps high-speed camera, convolving Ultra-High-Definition (UHD) sharp images with large-size kernels, adding defocus to images, and real-world blurry images captured by various camera models. Based on the MC-Blur dataset, we conduct extensive benchmarking studies to compare SOTA methods in different scenarios, analyze their efficiency, and investigate the buildataset’s capacity. These benchmarking results provide a comprehensive overview of the advantages and limitations of current deblurring methods, revealing our dataset’s advances. The dataset is available to the public athttps://github.com/HDCVLab/MC-Blur-Dataset.

To promote the application of object detectors in real scenes, out-of-distribution object detection (OOD-OD) is proposed to distinguish whether detected objects belong to the ones that are unseen during training or not. One of the key challenges is that detectors lack unknown data for supervision, and as a result, can produce overconfident detection results on OOD data. Thus, this task requires to synthesize OOD data for training, which achieves the goal of enhancing the ability of localizing and discriminating OOD objects. In this paper, we propose a novel method, i.e., PCA-Driven dynamic prototype enhancement, to explore exploiting Principal Component Analysis (PCA) to extract simulative OOD data for training and obtain dynamic prototypes that are related to the current input and are helpful for boosting the discrimination ability. Concretely, the last few principal components of the backbone features are utilized to calculate an OOD map that involves plentiful information that deviates from the correlation distribution of the input. The OOD map is further used to extract simulative OOD data for training, which alleviates the impact of lacking unknown data. Besides, for in-distribution (ID) data, the category-level semantic information of objects between the backbone features and the high-level features should be kept consistent. To this end, we utilize the residual principal components to extract dynamic prototypes that reflect the semantic information of the current backbone features. Next, we define a contrastive loss to leverage these prototypes to enlarge the semantic gap between the simulative OOD data and the features from the residual principal components, which improves the ability of discriminating OOD objects. In the experiments, we separately verify our method on OOD-OD and incremental object detection. The significant performance gains demonstrate the superiorities of our method.

Current oriented object detection methods mainly utilize a vanilla coordinate-angle representation for bounding box regression, which usually suffers from inconsistency between the bounding box regression losses and prediction errors induced with respect to different rotation angles, aspect ratios, and scales. Therefore, although the existing oriented object detectors have achieved very good performances under coarse evaluation metrics such as AP50, their performance significantly degrades when using stricter evaluation metric such as AP75. To address the abovementioned issues, we propose a new regression method with bounding box vectorization that implicitly represents the shape and orientation of an object with a set of orthogonal vectors. By doing this, the proposed method delicately avoids the inconsistency issues encountered in oriented bounding box regression. During training, we introduce the Tanimoto coefficient to evaluate the similarity of the bounding box vector in a shape- and orientation-aware manner, and we refer to the proposed box-to-vector loss as the B2V loss. In addition to 2D object detection, the proposed method can be easily generalized to 3D scenarios involving orientation estimation, such as autonomous driving. We evaluate the proposed method through extensive experiments conducted on four popular oriented object detection datasets, including both 2D and 3D datasets, where the proposed method significantly outperforms the recently developed state-of-the-art methods when using a more accurate evaluation metric.

We propose a novel master-slave architecture to solve the top- combinatorial multiarmed bandits (CMABs) problem with nonlinear bandit feedback and diversity constraints, which, to the best of our knowledge, is the first combinatorial bandits setting considering diversity constraints under bandit feedback. Specifically, to efficiently explore the combinatorial and constrained action space, we introduce six slave models with distinguished merits to generate diversified samples well balancing rewards and constraints as well as efficiency. Moreover, we propose teacher learning-based optimization and the policy cotraining technique to boost the performance of the multiple slave models. The master model then collects the elite samples provided by the slave models and selects the best sample estimated by a neural contextual UCB-based network (NeuralUCB) to decide on a tradeoff between exploration and exploitation. Thanks to the elaborate design of slave models, the cotraining mechanism among slave models, and the novel interactions between the master and slave models, our approach significantly surpasses existing state-of-the-art algorithms in both synthetic and real datasets for recommendation tasks. The code is available at https://github.com/huanghanchi/Master-slave-Algorithm-for-Top-K-Bandits.

Integrating adaptive learning rate and momentum techniques into stochastic gradient descent (SGD) leads to a large class of efficiently accelerated adaptive stochastic algorithms, such as AdaGrad, RMSProp, Adam, AccAdaGrad, and so on. In spite of their effectiveness in practice, there is still a large gap in their theories of convergences, especially in the difficult nonconvex stochastic setting. To fill this gap, we propose weighted AdaGrad with unified momentum and dubbed AdaUSM, which has the main characteristics that: 1) it incorporates a unified momentum scheme that covers both the heavy ball (HB) momentum and the Nesterov accelerated gradient (NAG) momentum and 2) it adopts a novel weighted adaptive learning rate that can unify the learning rates of AdaGrad, AccAdaGrad, Adam, and RMSProp. Moreover, when we take polynomially growing weights in AdaUSM, we obtain its O(log(T)/√T) convergence rate in the nonconvex stochastic setting. We also show that the adaptive learning rates of Adam and RMSProp correspond to taking exponentially growing weights in AdaUSM, thereby providing a new perspective for understanding Adam and RMSProp. Finally, comparative experiments of AdaUSM against SGD with momentum, AdaGrad, AdaEMA, Adam, and AMSGrad on various deep learning models and datasets are also carried out.

Data-dependent hashing methods aim to learn hash functions from the pairwise or triplet relationships among the data, which often lead to low efficiency and low collision rate by only capturing the local distribution of the data. To solve the limitation, we propose central similarity, in which the hash codes of similar data pairs are encouraged to approach a common center and those of dissimilar pairs to converge to different centers. As a new global similarity metric, central similarity can improve the efficiency and retrieval accuracy of hash learning. By introducing a new concept, hash centers, we principally formulate the computation of the proposed central similarity metric, in which the hash centers refer to a set of points scattered in the Hamming space with a sufficient mutual distance between each other. To construct well-separated hash centers, we provide two efficient methods: 1) leveraging the Hadamard matrix and Bernoulli distributions to generate data-independent hash centers and 2) learning data-dependent hash centers from data representations. Based on the proposed similarity metric and hash centers, we propose central similarity quantization (CSQ) that optimizes the central similarity between data points with respect to their hash centers instead of optimizing the local similarity to generate a high-quality deep hash function. We also further improve the CSQ with data-dependent hash centers, dubbed as CSQ with learnable center (CSQLC). The proposed CSQ and CSQLC are generic and applicable to image and video hashing scenarios. We conduct extensive experiments on large-scale image and video retrieval tasks, and the proposed CSQ yields noticeably boosted retrieval performance, i.e., 3%-20% in mean average precision (mAP) over the previous state-of-the-art methods, which also demonstrates that our methods can generate cohesive hash codes for similar data pairs and dispersed hash codes for dissimilar pairs.

AI-aided drug discovery (AIDD) is gaining popularity due to its potential to make the search for new pharmaceuticals faster, less expensive, and more effective. Despite its extensive use in numerous fields (e.g., ADMET prediction, virtual screening), little research has been conducted on the out-of-distribution (OOD) learning problem with noise. We present DrugOOD, a systematic OOD dataset curator and benchmark for AIDD. Particularly, we focus on the drug-target binding affinity prediction problem, which involves both macromolecule (protein target) and small-molecule (drug compound). DrugOOD offers an automated dataset curator with user-friendly customization scripts, rich domain annotations aligned with biochemistry knowledge, realistic noise level annotations, and rigorous benchmarking of SOTA OOD algorithms, as opposed to only providing fixed datasets. Since the molecular data is often modeled as irregular graphs using graph neural network (GNN) backbones, DrugOOD also serves as a valuable testbed for graph OOD learning problems. Extensive empirical studies have revealed a significant performance gap between in-distribution and out-of-distribution experiments, emphasizing the need for the development of more effective schemes that permit OOD generalization under noise for AIDD.

VQA is an ambitious task aiming to answer any image-related question. However, in reality, it is hard to build such a system once for all since the needs of users are continuously updated, and the system has to implement new functions. Thus, Continual Learning (CL) ability is a must in developing advanced VQA systems. Recently, a pioneer work split a VQA dataset into disjoint answer sets to study this topic. However, CL on VQA involves not only the expansion of label sets (new Answer sets). It is crucial to study how to answer questions when deploying VQA systems to new environments (new Visual scenes) and how to answer questions requiring new functions (new Question types). Thus, we propose CLOVE, a benchmark for Continual Learning On Visual quEstion answering, which contains scene- and function-incremental settings for the two aforementioned CL scenarios. In terms of methodology, the main difference between CL on VQA and classification is that the former additionally involves expanding and preventing forgetting of reasoning mechanisms, while the latter focusing on class representation. Thus, we propose a real-data-free replay-based method tailored for CL on VQA, named Scene Graph as Prompt for Symbolic Replay. Using a piece of scene graph as a prompt, it replays pseudo scene graphs to represent the past images, along with correlated QA pairs. A unified VQA model is also proposed to utilize the current and replayed data to enhance its QA ability. Finally, experimental results reveal challenges in CLOVE and demonstrate the effectiveness of our method. Code and data are available at https://github.com/showlab/CLVQA.

Deep neural network, despite its remarkable capability of discriminating targeted in-distribution samples, shows poor performance on detecting anomalous out-of-distribution data. To address this defect, state-of-the-art solutions choose to train deep networks on an auxiliary dataset of outliers. Various training criteria for these auxiliary outliers are proposed based on heuristic intuitions. However, we find that these intuitively designed outlier training criteria can hurt in-distribution learning and eventually lead to inferior performance. To this end, we identify three causes of the in-distribution incompatibility: contradictory gradient, false likelihood, and distribution shift. Based on our new understandings, we propose a new out-of-distribution detection method by adapting both the top-design of deep models and the loss function. Our method achieves in-distribution compatibility by pursuing less interference with the probabilistic characteristic of in-distribution features. On several benchmarks, our method not only achieves the state-of-the-art out-of-distribution detection performance but also improves the in-distribution accuracy.

Hao Sun, Zhexin Zhang, Fei Mi, Yasheng Wang, Wei Liu, Jianwei Cui, Bin Wang, Qun Liu, Minlie Huang. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers). 2023.

Regularization-based continuous learning (RCL) has been proven to be highly effective on struggling against catastrophic forgetting. However, its application in medical image segmentation (MIS) is relatively scarce. In this paper, we provide a unified description of 6 RCL methods (LwF, LwM, EWC, SI, MAS, and PIGWM) using Taylor expansion and investigate their performances in 2 classic MIS scenarios, namely retinal vessel segmentation (RVS) and cardiac left ventricle segmentation (CLVS) on 8 datasets (CHASE, DRHAGIS, RITE and STARE for the former and M&Ms, LVSC, ACDC and SCD for the latter). We also explore the influence of different task orders (easy to hard or hard to easy), optimizers (Adam or SGD), and parameter capacities (2, 3, 4 or 5 down- and up-sampling pairs) on the performance of these methods. Our experimental results show that these methods are capable of mitigating catastrophic forgetting to a certain extent. Comparing to a hard-to-easy order, most of the methods perform better on all of the already known tasks in an easy-to-hard order. Optimizer Adam performs better on RVS and CLVS. Capacity increases are obviously effective for CLVS, but they have no significant impact on RVS.

Cross-modal alignment is essential for vision-language pre-training (VLP) models to learn the correct corresponding information across different modalities. For this purpose, inspired by the success of masked language modeling (MLM) tasks in the NLP pre-training area, numerous masked modeling tasks have been proposed for VLP to further promote cross-modal interactions. The core idea of previous masked modeling tasks is to focus on reconstructing the masked tokens based on visible context for learning local-to-local alignment. However, most of them pay little attention to the global semantic features generated for the masked data, resulting in a limited cross-modal alignment ability of global representations. Therefore, in this paper, we propose a novel Semantic Completion Learning (SCL) task, complementary to existing masked modeling tasks, to facilitate global-to-local alignment. Specifically, the SCL task complements the missing semantics of masked data by capturing the corresponding information from the other modality, promoting learning more representative global features which have a great impact on the performance of downstream tasks. Moreover, we present a flexible vision encoder, which enables our model to perform image-text and video-text multimodal tasks simultaneously. Experimental results show that our proposed method obtains state-of-the-art performance on various vision-language benchmarks, such as visual question answering, image-text retrieval, and video-text retrieval.

This work studies the evaluation of explaining graph neural networks (GNNs), which is crucial to the credibility of post-hoc explainability in practical usage. Conventional evaluation metrics, and even explanation methods -- which mainly follow the paradigm of feeding the explanatory subgraph and measuring output difference -- always suffer from the notorious out-of-distribution (OOD) issue. In this work, we endeavor to confront the issue by introducing a novel evaluation metric, termed **O**OD-resistant **A**dversarial **R**obustness (OAR). Specifically, we draw inspiration from the notion of adversarial robustness and evaluate post-hoc explanation subgraphs by calculating their robustness under attack. On top of that, an elaborate OOD reweighting block is inserted into the pipeline to confine the evaluation process to the original data distribution. For applications involving large datasets, we further devise a **Sim**plified version of **OAR** (SimOAR), which achieves a significant improvement in computational efficiency at the cost of a small amount of performance. Extensive empirical studies validate the effectiveness of our OAR and SimOAR.

The adversarial attacks have attracted increasing attention in various fields including natural language processing. The current textual attacking models primarily focus on fooling models by adding character-/word-/sentence-level perturbations, ignoring their influence on human perception. In this paper, for the first time in the community, we propose a novel mode of textual attack, punctuation-level attack. With various types of perturbations, including insertion, displacement, deletion, and replacement, the punctuation-level attack achieves promising fooling rates against SOTA models on typical textual tasks and maintains minimal influence on human perception and understanding of the text by mere perturbation of single-shot single punctuation. Furthermore, we propose a search method named Text Position Punctuation Embedding and Paraphrase (TPPEP) to accelerate the pursuit of optimal position to deploy the attack, without exhaustive search, and we present a mathematical interpretation of TPPEP. Thanks to the integrated Text Position Punctuation Embedding (TPPE), the punctuation attack can be applied at a constant cost of time. Experimental results on public datasets and SOTA models demonstrate the effectiveness of the punctuation attack and the proposed TPPE. We additionally apply the single punctuation attack to summarization, semantic-similarity-scoring, and text-to-image tasks, and achieve encouraging results.

This paper investigates the sample complexity of learning a distributionally robust predictor under a particular distributional shift based on $\chi^2$-divergence, which is well known for its computational feasibility and statistical properties. We demonstrate that any hypothesis class $\mathcal{H}$ with finite VC dimension is distributionally robustly learnable. Moreover, we show that when the perturbation size is smaller than a constant, finite VC dimension is also necessary for distributionally robust learning by deriving a lower bound of sample complexity in terms of VC dimension.

We introduce CAMRL, the first curriculum-based asymmetric multi-task learning (AMTL) algorithm for dealing with multiple reinforcement learning (RL) tasks altogether. To mitigate the negative influence of customizing the one-off training order in curriculum-based AMTL, CAMRL switches its training mode between parallel single-task RL and asymmetric multi-task RL (MTRL), according to an indicator regarding the training time, the overall performance, and the performance gap among tasks. To leverage the multi-sourced prior knowledge flexibly and to reduce negative transfer in AMTL, we customize a composite loss with multiple differentiable ranking functions and optimize the loss through alternating optimization and the Frank-Wolfe algorithm. The uncertainty-based automatic adjustment of hyper-parameters is also applied to eliminate the need of laborious hyper-parameter analysis during optimization. By optimizing the composite loss, CAMRL predicts the next training task and continuously revisits the transfer matrix and network weights. We have conducted experiments on a wide range of benchmarks in multi-task RL, covering Gym-minigrid, Meta-world, Atari video games, vision-based PyBullet tasks, and RLBench, to show the improvements of CAMRL over the corresponding single-task RL algorithm and state-of-the-art MTRL algorithms. The code is available at: https://github.com/huanghanchi/CAMRL.

Pose transfer of human videos aims to generate a high-fidelity video of a target person imitating actions of a source person. A few studies have made great progress either through image translation with deep latent features or neural rendering with explicit 3D features. However, both of them rely on large amounts of training data to generate realistic results, and the performance degrades on more accessible Internet videos due to insufficient training frames. In this paper, we demonstrate that the dynamic details can be preserved even when trained from short monocular videos. Overall, we propose a neural video rendering framework coupled with an image-translation-based dynamic details generation network (D$^{2}$G-Net), which fully utilizes both the stability of explicit 3D features and the capacity of learning components. To be specific, a novel hybrid texture representation is presented to encode both the static and pose-varying appearance characteristics, which is then mapped to the image space and rendered as a detail-rich frame in the neural rendering stage. Through extensive comparisons, we demonstrate that our neural human video renderer is capable of achieving both clearer dynamic details and more robust performance even on accessible short videos with only 2 k$\sim$4 k frames, as illustrated in Fig. 1.

Continual learning (CL) aims to learn a non-stationary data distribution and not forget previous knowledge. The effectiveness of existing approaches that rely on memory replay can decrease over time as the model tends to overfit the stored examples. As a result, the model's ability to generalize well is significantly constrained. Additionally, these methods often overlook the inherent uncertainty in the memory data distribution, which differs significantly from the distribution of all previous data examples. To overcome these issues, we propose a principled memory evolution framework that dynamically adjusts the memory data distribution. This evolution is achieved by employing distributionally robust optimization (DRO) to make the memory buffer increasingly difficult to memorize. We consider two types of constraints in DRO: f-divergence and Wasserstein ball constraints. For f-divergence constraint, we derive a family of methods to evolve the memory buffer data in the continuous probability measure space with Wasserstein gradient flow (WGF). For Wasserstein ball constraint, we directly solve it in the euclidean space. Extensive experiments on existing benchmarks demonstrate the effectiveness of the proposed methods for alleviating forgetting. As a by-product of the proposed framework, our method is more robust to adversarial examples than compared CL methods.

Weakly supervised temporal action localization (WSTAL), which aims to locate the time interval of actions in an untrimmed video with only video-level action labels, has attracted increasing research interest in the past few years. However, a model trained with such labels will tend to focus on segments that contributions most to the video-level classification, leading to inaccurate and incomplete localization results. In this paper, we tackle the problem from a novel perspective of relation modeling and propose a method dubbed Bilateral Relation Distillation (BRD). The core of our method involves learning representations by jointly modeling the relation at the category and sequence levels. Specifically, category-wise latent segment representations are first obtained by different embedding networks, one for each category. We then distill knowledge obtained from a pre-trained language model to capture the category-level relations, which is achieved by performing correlation alignment and category-aware contrast in an intra- and inter-video manner. To model the relations among segments at the sequence-level, we elaborate a gradient-based feature augmentation method and encourage the learned latent representation of the augmented feature to be consistent with that of the original one. Extensive experiments illustrate that our approach achieves state-of-the-art results on THUMOS14 and ActivityNet1.3 datasets.

Face recognition has always been courted in computer vision and is especially amenable to situations with significant variations between frontal and profile faces. Traditional techniques make great strides either by synthesizing frontal faces from sizable datasets or by empirical pose invariant learning. In this paper, we propose a completely integrated embedded end-to-end Lie algebra residual architecture (LARNeXt) to achieve pose robust face recognition. First, we explore how the face rotation in the 3D space affects the deep feature generation process of convolutional neural networks (CNNs), and prove that face rotation in the image space is equivalent to an additive residual component in the feature space of CNNs, which is determined solely by the rotation. Second, on the basis of this theoretical finding, we further design three critical subnets to leverage a soft regression subnet with novel multi-fusion attention feature aggregation for efficient pose estimation, a residual subnet for decoding rotation information from input face images, and a gating subnet to learn rotation magnitude for controlling the strength of the residual component that contributes to the feature learning process. Finally, we conduct a large number of ablation experiments, and our quantitative and visualization results both corroborate the credibility of our theory and corresponding network designs. Our comprehensive experimental evaluations on frontal-profile face datasets, general unconstrained face recognition datasets, and industrial-grade tasks demonstrate that our method consistently outperforms the state-of-the-art ones.

Recent deep face hallucination methods show stunning performance in super-resolving severely degraded facial images, even surpassing human ability. However, these algorithms are mainly evaluated on non-public synthetic datasets. It is thus unclear how these algorithms perform on public face hallucination datasets. Meanwhile, most of the existing datasets do not well consider the distribution of races, which makes face hallucination methods trained on these datasets biased toward some specific races. To address the above two problems, in this paper, we build a public Ethnically Diverse Face dataset, EDFace-Celeb-1 M, and design a benchmark task for face hallucination. Our dataset includes 1.7 million photos that cover different countries, with relatively balanced race composition. To the best of our knowledge, it is the largest-scale and publicly available face hallucination dataset in the wild. Associated with this dataset, this paper also contributes various evaluation protocols and provides comprehensive analysis to benchmark the existing state-of-the-art methods. The benchmark evaluations demonstrate the performance and limitations of state-of-the-art algorithms. https://github.com/HDCVLab/EDFace-Celeb-1M.

Video deraining is an important task in computer vision as the unwanted rain hampers the visibility of videos and deteriorates the robustness of most outdoor vision systems. Despite the significant success which has been achieved for video deraining recently, two major challenges remain: 1) how to exploit the vast information among successive frames to extract powerful spatio-temporal features across both the spatial and temporal domains, and 2) how to restore high-quality derained videos with a high-speed approach. In this paper, we present a new end-to-end video deraining framework, dubbed Enhanced Spatio-Temporal Interaction Network (ESTINet), which considerably boosts current state-of-the-art video deraining quality and speed. The ESTINet takes the advantage of deep residual networks and convolutional long short-term memory, which can capture the spatial features and temporal correlations among successive frames at the cost of very little computational resource. Extensive experiments on three public datasets show that the proposed ESTINet can achieve faster speed than the competitors, while maintaining superior performance over the state-of-the-art methods. https://github.com/HDCVLab/Enhanced-Spatio-Temporal-Interaction-Learning-for-Video-Deraining.

Abstract Semantic parsing maps natural language questions into logical forms, which can be executed against a knowledge base for answers. In real-world applications, the performance of a parser is often limited by the lack of training data. To facilitate zero-shot learning, data synthesis has been widely studied to automatically generate paired questions and logical forms. However, data synthesis methods can hardly cover the diverse structures in natural languages, leading to a large gap in sentence structure between synthetic and natural questions. In this paper, we propose a decomposition-based method to unify the sentence structures of questions, which benefits the generalization to natural questions. Experiments demonstrate that our method significantly improves the semantic parser trained on synthetic data (+7.9% on KQA and +8.9% on ComplexWebQuestions in terms of exact match accuracy). Extensive analysis demonstrates that our method can better generalize to natural questions with novel text expressions compared with baselines. Besides semantic parsing, our idea potentially benefits other semantic understanding tasks by mitigating the distracting structure features. To illustrate this, we extend our method to the task of sentence embedding learning, and observe substantial improvements on sentence retrieval (+13.1% for Hit@1).

Single Image Super-Resolution (SISR) tasks have achieved significant performance with deep neural networks. However, the large number of parameters in CNN-based methods for SISR tasks require heavy computations. Although several efficient SISR models have been recently proposed, most are handcrafted and thus lack flexibility. In this work, we propose a novel differentiable Neural Architecture Search (NAS) approach on both the cell-level and network-level to search for lightweight SISR models. Specifically, the cell-level search space is designed based on an information distillation mechanism, focusing on the combinations of lightweight operations and aiming to build a more lightweight and accurate SR structure. The network-level search space is designed to consider the feature connections among the cells and aims to find which information flow benefits the cell most to boost the performance. Unlike the existing Reinforcement Learning (RL) or Evolutionary Algorithm (EA) based NAS methods for SISR tasks, our search pipeline is fully differentiable, and the lightweight SISR models can be efficiently searched on both the cell-level and network-level jointly on a single GPU. Experiments show that our methods can achieve state-of-the-art performance on the benchmark datasets in terms of PSNR, SSIM, and model complexity with merely 68G Multi-Adds for$\times 2$and 18G Multi-Adds for$\times 4$SR tasks.

Recently, numerous unsupervised cross-modal hashing methods have been proposed to deal the image-text retrieval tasks for the unlabeled cross-modal data. However, when these methods learn to generate hash codes, almost all of them lack modality-interaction in the following two aspects: 1) The instance similarity matrix used to guide the hashing networks training is constructed without image-text interaction, which fails to capture the fine-grained cross-modal cues to elaborately characterize the intrinsic semantic similarity among the datapoints. 2) The binary codes used for quantization loss are inferior because they are generated by directly quantizing a simple combination of continuous hash codes from different modalities without the interaction among these continuous hash codes. Such problems will cause the generated hash codes to be of poor quality and degrade the retrieval performance. Hence, in this paper, we propose a novel Unsupervised Cross-modal Hashing with Modality-interaction, termed UCHM. Specifically, by optimizing a novel hash-similarity-friendly loss, a modality-interaction-enabled (MIE) similarity generator is first trained to generate a superior MIE similarity matrix for the training set. Then, the generated MIE similarity matrix is utilized as guiding information to train the deep hashing networks. Furthermore, during the process of training the hashing networks, a novel bit-selection module is proposed to generate high-quality unified binary codes for the quantization loss with the interaction among continuous codes from different modalities, thereby further enhancing the retrieval performance. Extensive experiments on two widely used datasets show that the proposed UCHM outperforms state-of-the-art techniques on cross-modal retrieval tasks.

Exploiting fine-grained correspondence and visual-semantic alignments has shown great potential in image-text matching. Generally, recent approaches first employ a cross-modal attention unit to capture latent region-word interactions, and then integrate all the alignments to obtain the final similarity. However, most of them adopt one-time forward association or aggregation strategies with complex architectures or additional information, while ignoring the regulation ability of network feedback. In this paper, we develop two simple but quite effective regulators which efficiently encode the message output to automatically contextualize and aggregate cross-modal representations. Specifically, we propose (i) a Recurrent Correspondence Regulator (RCR) which facilitates the cross-modal attention unit progressively with adaptive attention factors to capture more flexible correspondence, and (ii) a Recurrent Aggregation Regulator (RAR) which adjusts the aggregation weights repeatedly to increasingly emphasize important alignments and dilute unimportant ones. Besides, it is interesting that RCR and RAR are "plug-and-play": both of them can be incorporated into many frameworks based on cross-modal interaction to obtain significant benefits, and their cooperation achieves further improvements. Extensive experiments on MSCOCO and Flickr30K datasets validate that they can bring an impressive and consistent R@1 gain on multiple models, confirming the general effectiveness and generalization ability of the proposed methods.

Matching-based methods, especially those based on space-time memory, are significantly ahead of other solutions in semi-supervised video object segmentation (VOS). However, continuously growing and redundant template features lead to an inefficient inference. To alleviate this, we propose a novel Sequential Weighted Expectation-Maximization (SWEM) network to greatly reduce the redundancy of memory features. Different from the previous methods which only detect feature redundancy between frames, SWEM merges both intra-frame and inter-frame similar features by leveraging the sequential weighted EM algorithm. Further, adaptive weights for frame features endow SWEM with the flexibility to represent hard samples, improving the discrimination of templates. Besides, the proposed method maintains a fixed number of template features in memory, which ensures the stable inference complexity of the VOS system. Extensive experiments on commonly used DAVIS and YouTube-VOS datasets verify the high efficiency (36 FPS) and high performance (84.3% J&F on DAVIS 2017 validation dataset) of SWEM.

In this paper, we propose a novel method to realize multimodal image registration and fusion in a mutually reinforcing framework, termed as RFNet. We handle the registration in a coarse-to-fine fashion. For the first time, we exploit the feedback of image fusion to promote the registration accuracy rather than treating them as two separate issues. The fine-registered results also improve the fusion performance. Specifically, for image registration, we solve the bottlenecks of defining registration metrics applicable for multi-modal images and facilitating the network convergence. The metrics are defined based on image translation and image fusion respectively in the coarse and fine stages. The convergence is facilitated by the designed metrics and a deformable convolution-based network. For image fusion, we focus on texture preservation, which not only increases the information amount and quality of fusion results but also improves the feedback of fusion results. The proposed method is evaluated on multi-modal images with large global parallaxes, images with local misalignments and aligned images to validate the performances of registration and fusion. The results in these cases demonstrate the effectiveness of our method.

Recently, MLP-like vision models have achieved promising performances on mainstream visual recognition tasks. In contrast with vision transformers and CNNs, the success of MLP-like models shows that simple information fusion operations among tokens and channels can yield a good representation power for deep recognition models. However, existing MLP-like models fuse tokens through static fusion operations, lacking adaptability to the contents of the tokens to be mixed. Thus, customary information fusion procedures are not effective enough. To this end, this paper presents an efficient MLP-like network architecture, dubbed DynaMixer, resorting to dynamic information fusion. Critically, we propose a procedure, on which the DynaMixer model relies, to dynamically generate mixing matrices by leveraging the contents of all the tokens to be mixed. To reduce the time complexity and improve the robustness, a dimensionality reduction technique and a multi-segment fusion mechanism are adopted. Our proposed DynaMixer model (97M parameters) achieves 84.3% top-1 accuracy on the ImageNet-1K dataset without extra training data, performing favorably against the state-of-the-art vision MLP models. When the number of parameters is reduced to 26M, it still achieves 82.7% top-1 accuracy, surpassing the existing MLP-like models with a similar capacity. The code is available at \url{https://github.com/ziyuwwang/DynaMixer}.

Image-text retrieval is an essential task of information retrieval, in which the models with the Vision-and-Language Pretraining(VLP) are able to achieve ideal accuracy compared with the ones without VLP. Among different VLP approaches, the single-stream models achieve the overall best retrieval accuracy, but slower inference speed. Recently, researchers have introduced the two-stage retrieval setting commonly used in the information retrieval field to the single-stream VLP model for a better accuracy/efficiency trade-off. However, the retrieval accuracy and efficiency are still unsatisfactory mainly due to the limitations of the patch-based visual unimodal encoder in these VLP models. The unimodal encoders are trained on pure visual data, so the visual features extracted by them are difficult to align with the textual features and it is also difficult for the multi-modal encoder to understand visual information. Under these circumstances, we propose an accurate and efficient two-stage image-text retrieval model via Contrastive Alignment and visual Contexts modeling(CAliC). In the first stage of the proposed model, the visual unimodal encoder is pretrained with cross-modal contrastive learning to extract easily aligned visual features, which improves the retrieval accuracy and the inference speed. In the second stage of the proposed model, we introduce a new visual contexts modeling task during pretraining to help the multi-modal encoder better understand the visual information and get more accurate predictions. Extensive experimental evaluation validates the effectiveness of our proposed approach, which achieves a higher retrieval accuracy while keeping a faster inference speed, and outperforms existing state-of-the-art retrieval methods on image-text retrieval tasks over Flickr30K and COCO benchmarks.

Video-Language Pretraining (VLP), which aims to learn transferable representation to advance a wide range of video-text downstream tasks, has recently received increasing attention. Best performing works rely on large-scale, 3rd-person video-text datasets, such as HowTo100M. In this work, we exploit the recently released Ego4D dataset to pioneer Egocentric VLP along three directions. (i) We create EgoClip, a 1st-person video-text pretraining dataset comprising 3.8M clip-text pairs well-chosen from Ego4D, covering a large variety of human daily activities. (ii) We propose a novel pretraining objective, dubbed EgoNCE, which adapts video-text contrastive learning to the egocentric domain by mining egocentric-aware positive and negative samples. (iii) We introduce EgoMCQ, a development benchmark that is close to EgoClip and hence can support effective validation and fast exploration of our design decisions in EgoClip and EgoNCE. Furthermore, we demonstrate strong performance on five egocentric downstream tasks across three datasets: video-text retrieval on EPIC-KITCHENS-100; action recognition on Charades-Ego; natural language query, moment query, and object state change classification on Ego4D challenge benchmarks. The dataset and code are available at https://github.com/showlab/EgoVLP.

Adam is one of the most influential adaptive stochastic algorithms for training deep neural networks, which has been pointed out to be divergent even in the simple convex setting via a few simple counterexamples. Many attempts, such as decreasing an adaptive learning rate, adopting a big batch size, incorporating a temporal decorrelation technique, seeking an analogous surrogate, etc., have been tried to promote Adam-type algorithms to converge. In contrast with existing approaches, we introduce an alternative easy-to-check sufficient condition, which merely depends on the parameters of the base learning rate and combinations of historical second-order moments, to guarantee the global convergence of generic Adam for solving large-scale non-convex stochastic optimization. This observation, coupled with this sufficient condition, gives much deeper interpretations on the divergence of Adam. On the other hand, in practice, mini-Adam and distributed-Adam are widely used without any theoretical guarantee. We further give an analysis on how the batch size or the number of nodes in the distributed system affects the convergence of Adam, which theoretically shows that mini-batch and distributed Adam can be linearly accelerated by using a larger mini-batch size or a larger number of nodes. At last, we apply the generic Adam and mini-batch Adam with the sufficient condition for solving the counterexample and training several neural networks on various real-world datasets. Experimental results are exactly in accord with our theoretical analysis.

With the recent advancement of deep convolutional neural networks, significant progress has been made in general face recognition. However, the state-of-the-art general face recognition models do not generalize well to occluded face images, which are exactly the common cases in real-world scenarios. The potential reasons are the absences of large-scale occluded face data for training and specific designs for tackling corrupted features brought by occlusions. This article presents a novel face recognition method that is robust to occlusions based on a single end-to-end deep neural network. Our approach, named FROM (Face Recognition with Occlusion Masks), learns to discover the corrupted features from the deep convolutional neural networks, and clean them by the dynamically learned masks. In addition, we construct massive occluded face images to train FROM effectively and efficiently. FROM is simple yet powerful compared to the existing methods that either rely on external detectors to discover the occlusions or employ shallow models which are less discriminative. Experimental results on the LFW, Megaface challenge 1, RMF2, AR dataset and other simulated occluded/masked datasets confirm that FROM dramatically improves the accuracy under occlusions, and generalizes well on general face recognition.

Human motion prediction aims to generate future motions based on the observed human motions. Witnessing the success of Recurrent Neural Networks (RNN) in modeling sequential data, recent works utilize RNNs to model human-skeleton motions on the observed motion sequence and predict future human motions. However, these methods disregard the existence of the spatial coherence among joints and the temporal evolution among skeletons, which reflects the crucial characteristics of human motions in spatiotemporal space. To this end, we propose a novel Skeleton-Joint Co-Attention Recurrent Neural Networks (SC-RNN) to capture the spatial coherence among joints, and the temporal evolution among skeletons simultaneously on a skeleton-joint co-attention feature map in spatiotemporal space. First, a skeleton-joint feature map is constructed as the representation of the observed motion sequence. Second, we design a new Skeleton-Joint Co-Attention (SCA) mechanism to dynamically learn a skeleton-joint co-attention feature map of this skeleton-joint feature map, which can refine the useful observed motion information to predict one future motion. Third, a variant of GRU embedded with SCA collaboratively models the human-skeleton motion and human-joint motion in spatiotemporal space by regarding the skeleton-joint co-attention feature map as the motion context. Experimental results of human motion prediction demonstrate that the proposed method outperforms the competing methods.

Generative Adversarial Networks (GANs) are formulated as minimax game problems that generative networks attempt to approach real data distributions by adversarial learning against discriminators which learn to distinguish generated samples from real ones, of which the intrinsic problem complexity poses challenges to performance and robustness. In this work, we aim to boost model learning from the perspective of network architectures, by incorporating recent progress on automated architecture search into GANs. Specially we propose a fully differentiable search framework, dubbedalphaGAN, where the searching process is formalized as solving a bi-level minimax optimization problem. The outer-level objective aims for seeking an optimal network architecture towards pure Nash Equilibrium conditioned on the network parameters of generators and discriminators optimized with a traditional adversarial loss within inner level. The entire optimization performs a first-order approach by alternately minimizing the two-level objective in a fully differentiable manner that enables obtaining a suitable architecture efficiently from an enormous search space. Extensive experiments on CIFAR-10 and STL-10 datasets show that our algorithm can obtain high-performing architectures only with 3-GPU hours on a single GPU in the search space comprised of approximate$2\times 10^{11}$possible configurations. We further validate the method on the state-of-the-art network StyleGAN2, and push the score of Fréchet Inception Distance (FID) further, i.e., achieving 1.94 on CelebA, 2.86 on LSUN-church and 2.75 on FFHQ, with relative improvements$3\%{\sim} 26\%$over the baseline architecture. We also provide a comprehensive analysis of the behavior of the searching process and the properties of searched architectures, which would benefit further research on architectures for generative models. Codes and models are available athttps://github.com/yuesongtian/AlphaGAN.

This paper proposes a novel pretext task to address the self-supervised video representation learning problem. Specifically, given an unlabeled video clip, we compute a series of spatio-temporal statistical summaries, such as the spatial location and dominant direction of the largest motion, the spatial location and dominant color of the largest color diversity along the temporal axis, etc. Then a neural network is built and trained to yield the statistical summaries given the video frames as inputs. In order to alleviate the learning difficulty, we employ several spatial partitioning patterns to encode rough spatial locations instead of exact spatial Cartesian coordinates. Our approach is inspired by the observation that human visual system is sensitive to rapidly changing contents in the visual field, and only needs impressions about rough spatial locations to understand the visual contents. To validate the effectiveness of the proposed approach, we conduct extensive experiments with four 3D backbone networks, i.e., C3D, 3D-ResNet, R(2+1)D and S3D-G. The results show that our approach outperforms the existing approaches across these backbone networks on four downstream video analysis tasks including action recognition, video retrieval, dynamic scene recognition, and action similarity labeling. The source code is publicly available at: https://github.com/laura-wang/video_repres_sts.

Temporal sentence grounding in videos aims to localize one target video segment, which semantically corresponds to a given sentence. Unlike previous methods mainly focusing on matching semantics between the sentence and different video segments, in this paper, we propose a novel semantic conditioned dynamic modulation (SCDM) mechanism, which leverages the sentence semantics to modulate the temporal convolution operations for better correlating and composing the sentence-relevant video contents over time. The proposed SCDM also performs dynamically with respect to the diverse video contents so as to establish a precise semantic alignment between sentence and video. By coupling the proposed SCDM with a hierarchical temporal convolutional architecture, video segments with various temporal scales are composed and localized. Besides, more fine-grained clip-level actionness scores are also predicted with the SCDM-coupled temporal convolution on the bottom layer of the overall architecture, which are further used to adjust the temporal boundaries of the localized segments and thereby lead to more accurate grounding results. Experimental results on benchmark datasets demonstrate that the proposed model can improve the temporal grounding accuracy consistently, and further investigation experiments also illustrate the advantages of SCDM on stabilizing the model training and associating relevant video contents for temporal sentence grounding. Our code for this paper is available at https://github.com/yytzsy/SCDM-TPAMI.

In real-world scenarios, many factors may harm face recognition performance,e.g., large pose, bad illumination, low resolution, blur and noise. To address these challenges, previous efforts usually first restore the low-quality faces to high-quality ones and then perform face recognition. However, most of these methods are stage-wise, which is sub-optimal and deviates from the reality. In this paper, we address all these challenges jointly for unconstrained face recognition. We propose anMulti-DegradationFaceRestoration (MDFR) model to restore frontalized high-quality faces from the given low-quality ones under arbitrary facial poses, with three distinct novelties. First, MDFR is a well-designed encoder-decoder architecture which extracts feature representation from an input face image with arbitrary low-quality factors and restores it to a high-quality counterpart. Second, MDFR introduces a pose residual learning strategy along with a 3D-basedPoseNormalizationModule (PNM), which can perceive the pose gap between the input initial pose and its real-frontal pose to guide the face frontalization. Finally, MDFR can generate frontalized high-quality face images by a single unified network, showing a strong capability of preserving face identity. Qualitative and quantitative experiments on both controlled and in-the-wild benchmarks demonstrate the superiority of MDFR over state-of-the-art methods on both face frontalization and face restoration.

We present a versatile model, FaceAnime, for various video generation tasks from still images. Video generation from a single face image is an interesting problem and usually tackled by utilizing Generative Adversarial Networks (GANs) to integrate information from the input face image and a sequence of sparse facial landmarks. However, the generated face images usually suffer from quality loss, image distortion, identity change, and expression mismatching due to the weak representation capacity of the facial landmarks. In this paper, we propose to “imagine” a face video from a single face image according to the reconstructed 3D face dynamics, aiming to generate a realistic and identity-preserving face video, with precisely predicted pose and facial expression. The 3D dynamics reveal changes of the facial expression and motion, and can serve as a strong prior knowledge for guiding highly realistic face video generation. In particular, we explore face video prediction and exploit a well-designed 3D dynamic prediction network to predict a 3D dynamic sequence for a single face image. The 3D dynamics are then further rendered by the sparse texture mapping algorithm to recover structural details and sparse textures for generating face frames. Our model is versatile for various AR/VR and entertainment applications, such as face video retargeting and face video prediction. Superior experimental results have well demonstrated its effectiveness in generating high-fidelity, identity-preserving, and visually pleasant face video clips from a single source face image.

Visual tracking is typically solved as a discriminative learning problem that usually requires high-quality samples for online model adaptation. It is a critical and challenging problem to evaluate the training samples collected from previous predictions and employ sample selection by their quality to train the model. To tackle the above problem, we propose a joint discriminative learning scheme with the progressive multistage optimization policy of sample selection for robust visual tracking. The proposed scheme presents a novel time-weighted and detection-guided self-paced learning strategy for easy-to-hard sample selection, which is capable of tolerating relatively large intraclass variations while maintaining interclass separability. Such a self-paced learning strategy is jointly optimized in conjunction with the discriminative tracking process, resulting in robust tracking results. Experiments on the benchmark datasets demonstrate the effectiveness of the proposed learning framework.

Deep learning-based object detection and instance segmentation have achieved unprecedented progress. In this article, we propose complete-IoU (CIoU) loss and Cluster-NMS for enhancing geometric factors in both bounding-box regression and nonmaximum suppression (NMS), leading to notable gains of average precision (AP) and average recall (AR), without the sacrifice of inference efficiency. In particular, we consider three geometric factors, that is: 1) overlap area; 2) normalized central-point distance; and 3) aspect ratio, which are crucial for measuring bounding-box regression in object detection and instance segmentation. The three geometric factors are then incorporated into CIoU loss for better distinguishing difficult regression cases. The training of deep models using CIoU loss results in consistent AP and AR improvements in comparison to widely adopted$\ell _{n}$-norm loss and IoU-based loss. Furthermore, we propose Cluster-NMS, where NMS during inference is done by implicitly clustering detected boxes and usually requires fewer iterations. Cluster-NMS is very efficient due to its pure GPU implementation, and geometric factors can be incorporated to improve both AP and AR. In the experiments, CIoU loss and Cluster-NMS have been applied to state-of-the-art instance segmentation (e.g., YOLACT and BlendMask-RT), and object detection (e.g., YOLO v3, SSD, and Faster R-CNN) models. Taking YOLACT on MS COCO as an example, our method achieves performance gains as +1.7 AP and +6.2 AR100for object detection, and +1.1 AP and +3.5 AR100for instance segmentation, with 27.1 FPS on one NVIDIA GTX 1080Ti GPU. All the source code and trained models are available athttps://github.com/Zzh-tju/CIoU.

Few-shot learning, aiming to learn novel concepts from one or a few labeled examples, is an interesting and very challenging problem with many practical advantages. Existing few-shot methods usually utilize data of the same classes to train the feature embedding module and in a row, which is unable to learn adapting to new tasks. Besides, traditional few-shot models fail to take advantage of the valuable relations of the support-query pairs, leading to performance degradation. In this article, we propose a transductive relation-propagation graph neural network (GNN) with a decoupling training strategy (TRPN-D) to explicitly model and propagate such relations across support-query pairs, and empower the few-shot module the ability of transferring past knowledge to new tasks via the decoupling training. Our few-shot module, namely TRPN, treats the relation of each support-query pair as a graph node, named relational node, and resorts to the known relations between support samples, including both intraclass commonality and interclass uniqueness. Through relation propagation, the model could generate the discriminative relation embeddings for support-query pairs. To the best of our knowledge, this is the first work that decouples the training of the embedding network and the few-shot graph module with different tasks, which might offer a new way to solve the few-shot learning problem. Extensive experiments conducted on several benchmark datasets demonstrate that our method can significantly outperform a variety of state-of-the-art few-shot learning methods.

MoCo [11] is effective for unsupervised image representation learning. In this paper, we propose VideoMoCo for unsupervised video representation learning. Given a video sequence as an input sample, we improve the temporal feature representations of MoCo from two perspectives. First, we introduce a generator to drop out several frames from this sample temporally. The discriminator is then learned to encode similar feature representations regardless of frame removals. By adaptively dropping out different frames during training iterations of adversarial learning, we augment this input sample to train a tempo-rally robust encoder. Second, we use temporal decay to model key attenuation in the memory queue when computing the contrastive loss. As the momentum encoder updates after keys enqueue, the representation ability of these keys degrades when we use the current input sample for contrastive learning. This degradation is reflected via temporal decay to attend the input sample to recent keys in the queue. As a result, we adapt MoCo to learn video representations without empirically designing pretext tasks. By empowering the temporal robustness of the encoder and modeling the temporal decay of the keys, our VideoMoCo improves MoCo temporally based on contrastive learning. Experiments on benchmark datasets including UCF101 and HMDB51 show that VideoMoCo stands as a state-of-the-art video representation learning method.

Controllable Image Captioning (CIC) — generating image descriptions following designated control signals — has received unprecedented attention over the last few years. To emulate the human ability in controlling caption generation, current CIC studies focus exclusively on control signals concerning objective properties, such as contents of interest or descriptive patterns. However, we argue that almost all existing objective control signals have overlooked two indispensable characteristics of an ideal control signal: 1) Event-compatible: all visual contents referred to in a single sentence should be compatible with the described activity. 2) Sample-suitable: the control signals should be suitable for a specific image sample. To this end, we propose a new control signal for CIC: Verb-specific Semantic Roles (VSR). VSR consists of a verb and some semantic roles, which represents a targeted activity and the roles of entities involved in this activity. Given a designated VSR, we first train a grounded semantic role labeling (GSRL) model to identify and ground all entities for each role. Then, we propose a semantic structure planner (SSP) to learn human-like descriptive semantic structures. Lastly, we use a role-shift captioning model to generate the captions. Extensive experiments and ablations demonstrate that our framework can achieve better controllability than several strong base-lines on two challenging CIC benchmarks. Besides, we can generate multi-level diverse captions easily. The code is available at: https://github.com/mad-red/VSR-guided-CIC.

Universal style transfer retains styles from reference images in content images. While existing methods have achieved state-of-the-art style transfer performance, they are not aware of the content leak phenomenon that the image content may corrupt after several rounds of stylization process. In this paper, we propose ArtFlow to prevent content leak during universal style transfer. ArtFlow consists of reversible neural flows and an unbiased feature transfer module. It supports both forward and backward inferences and operates in a projection-transfer-reversion scheme. The forward inference projects input images into deep features, while the backward inference remaps deep features back to input images in a lossless and unbiased way. Extensive experiments demonstrate that ArtFlow achieves comparable performance to state-of-the-art style transfer methods while avoiding content leak.

Image virtual try-on replaces the clothes on a person image with a desired in-shop clothes image. It is challenging because the person and the in-shop clothes are unpaired. Existing methods formulate virtual try-on as either in-painting or cycle consistency. Both of these two formulations encourage the generation networks to reconstruct the input image in a self-supervised manner. However, existing methods do not differentiate clothing and non-clothing regions. A straightforward generation impedes the virtual try-on quality because of the heavily coupled image contents. In this paper, we propose a Disentangled Cycle-consistency Try-On Network (DCTON). The DCTON is able to produce highly-realistic try-on images by disentangling important components of virtual try-on including clothes warping, skin synthesis, and image composition. Moreover, DCTON can be naturally trained in a self-supervised manner following cycle consistency learning. Extensive experiments on challenging benchmarks show that DCTON outperforms state-of-the-art approaches favorably.

Image virtual try-on aims to fit a garment image (target clothes) to a person image. Prior methods are heavily based on human parsing. However, slightly-wrong segmentation results would lead to unrealistic try-on images with large artifacts. A recent pioneering work employed knowledge distillation to reduce the dependency of human parsing, where the try-on images produced by a parser-based method are used as supervisions to train a "student" network without relying on segmentation, making the student mimic the try-on ability of the parser-based model. However, the image quality of the student is bounded by the parser-based model. To address this problem, we propose a novel approach, "teacher-tutor-student" knowledge distillation, which is able to produce highly photo-realistic images without human parsing, possessing several appealing advantages compared to prior arts. (1) Unlike existing work, our approach treats the fake images produced by the parser-based method as "tutor knowledge", where the artifacts can be corrected by real "teacher knowledge", which is extracted from the real person images in a self-supervised way. (2) Other than using real images as supervisions, we formulate knowledge distillation in the try-on problem as distilling the appearance flows between the person image and the garment image, enabling us to find accurate dense correspondences between them to produce high-quality results. (3) Extensive evaluations show large superiority of our method (see Fig. 1).

User-intended visual content fills the hole regions of an input image in the image editing scenario. The coarse low- level inputs, which typically consist of sparse sketch lines and color dots, convey user intentions for content creation (i.e., free-form editing). While existing methods combine an input image and these low-level controls for CNN inputs, the corresponding feature representations are not sufficient to convey user intentions, leading to unfaithfully generated content. In this paper, we propose DeFLOCNet which relies on a deep encoder-decoder CNN to retain the guidance of these controls in the deep feature representations. In each skip-connection layer, we design a structure generation block. Instead of attaching low-level controls to an input image, we inject these controls directly into each structure generation block for sketch line refinement and color propagation in the CNN feature space. We then concatenate the modulated features with the original decoder features for structure generation. Meanwhile, DeFLOCNet involves another decoder branch for texture generation and detail enhancement. Both structures and textures are rendered in the decoder, leading to user-intended editing results. Experiments on benchmarks demonstrate that DeFLOCNet effectively transforms different user intentions to create visually pleasing content.

Current face forgery detection methods achieve high accuracy under the within-database scenario where training and testing forgeries are synthesized by the same algorithm. However, few of them gain satisfying performance under the cross-database scenario where training and testing forgeries are synthesized by different algorithms. In this paper, we find that current CNN-based detectors tend to overfit to method-specific color textures and thus fail to generalize. Observing that image noises remove color textures and expose discrepancies between authentic and tampered regions, we propose to utilize the high-frequency noises for face forgery detection. We carefully devise three functional modules to take full advantage of the high-frequency features. The first is the multi-scale high-frequency feature extraction module that extracts high-frequency noises at multiple scales and composes a novel modality. The second is the residual-guided spatial attention module that guides the low-level RGB feature extractor to concentrate more on forgery traces from a new perspective. The last is the cross-modality attention module that leverages the correlation between the two complementary modalities to promote feature learning for each other. Comprehensive evaluations on several benchmark databases corroborate the superior generalization performance of our proposed method.

Multi-scale and multi-patch deep models have been shown effective in removing blurs of dynamic scenes. However, these methods still suffer from one major obstacle: manually designing a lightweight and high-efficiency network is challenging and time-consuming. To tackle this obstacle, we propose a novel deblurring method, dubbed PyNAS (pyramid neural architecture search network), towards automatically designing hyper-parameters including the scales, patches, and standard cell operators. The proposed PyNAS adopts gradient-based search strategies and innovatively searches the hierarchy patch and scale scheme not limited to cell searching. Specifically, we introduce a hierarchical search strategy tailored to the multi-scale and multi-patch deblurring task. The strategy follows the principle that the first distinguishes between the top-level (pyramid-scales and pyramid-patches) and bottom-level variables (cell operators) and then searches multi-scale variables using the top-to-bottom principle. During the search stage, PyNAS employs an early stopping strategy to avoid the collapse and computational issues. Furthermore, we use a path-level binarization mechanism for multi-scale cell searching to save the memory consumption. Our primary contribution is a real-time deblurring algorithm (around 58 fps) for 720p images while achieves state-of-the-art deblurring performance on the GoPro and Video Deblurring datasets.

Recently, Cross-Modal Hamming space Retrieval (CMHR) regains ever-increasing attention, mainly benefiting from the excellent representation capability of deep neural networks. On the other hand, the vulnerability of deep networks exposes a deep cross-modal retrieval system to various safety risks (e.g., adversarial attack). However, attacking deep cross-modal Hamming retrieval remains underexplored. In this paper, we propose an effective Adversarial Attack on Deep Cross-Modal Hamming Retrieval, dubbed AACH, which fools a target deep CMHR model in a black-box setting. Specifically, given a target model, we first construct its substitute model to exploit cross-modal correlations within hamming space, with which we create adversarial examples by limitedly querying from a target model. Furthermore, to enhance the efficiency of adversarial attacks, we design a triplet construction module to exploit cross-modal positive and negative instances. In this way, perturbations can be learned to fool the target model through pulling perturbed examples far away from the positive instances whereas pushing them close to the negative ones. Extensive experiments on three widely used cross-modal (image and text) retrieval benchmarks demonstrate the superiority of the proposed AACH. We find that AACH can successfully attack a given target deep CMHR model with fewer interactions, and that its performance is on par with previous state-of-the-art attacks.

With the recent success of deep neural networks, remarkable progress has been achieved on face recognition. However, collecting large-scale real-world training data for face recognition has turned out to be challenging, especially due to the label noise and privacy issues. Meanwhile, existing face recognition datasets are usually collected from web images, lacking detailed annotations on attributes (e.g., pose and expression), so the influences of different attributes on face recognition have been poorly investigated. In this paper, we address the above-mentioned issues in face recognition using synthetic face images, i.e., SynFace. Specifically, we first explore the performance gap between recent state-of-the-art face recognition models trained with synthetic and real face images. We then analyze the underlying causes behind the performance gap, e.g., the poor intraclass variations and the domain gap between synthetic and real face images. Inspired by this, we devise the SynFace with identity mixup (IM) and domain mixup (DM) to mitigate the above performance gap, demonstrating the great potentials of synthetic data for face recognition. Furthermore, with the controllable face synthesis model, we can easily manage different factors of synthetic face generation, including pose, expression, illumination, the number of identities, and samples per identity. Therefore, we also perform a systematically empirical analysis on synthetic face images to provide some insights on how to effectively utilize synthetic data for face recognition.

Increasingly, modern mobile devices allow capturing images at Ultra-High-Definition (UHD) resolution, which includes 4K and 8K images. However, current single image super-resolution (SISR) methods focus on super-resolving images to ones with resolution up to high definition (HD) and ignore higher-resolution UHD images. To explore their performance on UHD images, in this paper, we first introduce two large-scale image datasets, UHDSR4K and UHDSR8K, to benchmark existing SISR methods. With 70,000 V100 GPU hours of training, we benchmark these methods on 4K and 8K resolution images under seven different settings to provide a set of baseline models. Moreover, we propose a baseline model, called Mesh Attention Network (MANet) for SISR. The MANet applies the attention mechanism in both different depths (horizontal) and different levels of receptive field (vertical). In this way, correlations among feature maps are learned, enabling the network to focus on more important features.

Most neural network pruning methods, such as filter-level and layer-level prunings, prune the network model along one dimension (depth, width, or resolution) solely to meet a computational budget. However, such a pruning policy often leads to excessive reduction of that dimension, thus inducing a huge accuracy loss. To alleviate this issue, we argue that pruning should be conducted along three dimensions comprehensively. For this purpose, our pruning framework formulates pruning as an optimization problem. Specifically, it first casts the relationships between a certain model’s accuracy and depth/width/resolution into a polynomial regression and then maximizes the polynomial to acquire the optimal values for the three dimensions. Finally, the model is pruned along the three optimal dimensions accordingly. In this framework, since collecting too much data for training the regression is very time-costly, we propose two approaches to lower the cost: 1) specializing the polynomial to ensure an accurate regression even with less training data; 2) employing iterative pruning and fine-tuning to collect the data faster. Extensive experiments show that our proposed algorithm surpasses state-of-the-art pruning algorithms and even neural architecture search-based algorithms.

Face recognition is an important yet challenging problem in computer vision. A major challenge in practical face recognition applications lies in significant variations between profile and frontal faces. Traditional techniques address this challenge either by synthesizing frontal faces or by pose invariant learning. In this paper, we propose a novel method with Lie algebra theory to explore how face rotation in the 3D space affects the deep feature generation process of convolutional neural networks (CNNs). We prove that face rotation in the image space is equivalent to an additive residual component in the feature space of CNNs, which is determined solely by the rotation. Based on this theoretical finding, we further design a Lie Algebraic Residual Network (LARNet) for tackling pose robust face recognition. Our LARNet consists of a residual subnet for decoding rotation information from input face images, and a gating subnet to learn rotation magnitude for controlling the strength of the residual component contributing to the feature learning process. Comprehensive experimental evaluations on both frontal-profile face datasets and general face recognition datasets convincingly demonstrate that our method consistently outperforms the state-of-the-art ones.

Few-shot object detection (FSOD) aims to detect new objects based on few annotated samples. To alleviate the impact of few samples, enhancing the generalization and discrimination abilities of detectors on new objects plays an important role. In this paper, we explore employing Singular Value Decomposition (SVD) to boost both the generalization and discrimination abilities. In specific, we propose a novel method, namely, SVD-Dictionary enhancement, to build two separated spaces based on the sorted singular values. Concretely, the eigenvectors corresponding to larger singular values are used to build the generalization space in which localization is performed, as these eigenvectors generally suppress certain variations (e.g., the variation of styles) and contain intrinsical characteristics of objects. Meanwhile, since the eigenvectors corresponding to relatively smaller singular values may contain richer category-related information, we can utilize them to build the discrimination space in which classification is performed. Dictionary learning is further leveraged to capture high-level discriminative information from the discrimination space, which is beneficial for improving detection accuracy. In the experiments, we separately verify the effectiveness of our method on PASCAL VOC and COCO benchmarks. Particularly, for the 2-shot case in VOC split1, our method significantly outperforms the baseline by 6.2\%. Moreover, visualization analysis shows that our method is instrumental in doing FSOD.

Recent Convolutional Neural Networks (CNNs) have achieved significant success by stacking multiple convolutional blocks, named procedures in this paper, to extract semantic features. However, they use the same procedure sequence for all inputs, regardless of the intermediate features.This paper proffers a simple yet effective idea of constructing parallel procedures and assigning similar intermediate features to the same specialized procedures in a divide-and-conquer fashion. It relieves each procedure's learning difficulty and thus leads to superior performance. Specifically, we propose a routing-by-memory mechanism for existing CNN architectures. In each stage of the network, we introduce parallel Procedural Units (PUs). A PU consists of a memory head and a procedure. The memory head maintains a summary of a type of features. For an intermediate feature, we search its closest memory and forward it to the corresponding procedure in both training and testing. In this way, different procedures are tailored to different features and therefore tackle them better.Networks with the proposed mechanism can be trained efficiently using a four-step training strategy. Experimental results show that our method improves VGGNet, ResNet, and EfficientNet's accuracies on Tiny ImageNet, ImageNet, and CIFAR-100 benchmarks with a negligible extra computational cost.

Abstract Previous deep learning methods usually required large‐scale annotated data, which is computationally exhaustive and unrealistic in certain scenarios. Therefore, few‐shot learning, where only a few annotated training images are available for training, has attracted increasing attention these days, showing huge potential in practical applications, such as portable equipment or security inspection, and so on. However, current few‐shot learning methods usually neglect the valuable semantic correlations between samples, thereby failing in extracting discriminating relations to achieve accurate predictive results. In this work, extending on a recent state‐of‐the‐art few‐shot learning method, transductive relation‐propagation network (TRPN), which considers the correlations between training samples, a constrained relation‐propagation network is proposed to further regularise the distilled correlations and thus achieve favourable few‐shot classification performance. The proposed framework contains three main components, namely preprocess module, relational propagation module, and relation constraint module. First, sample features are extracted and a relation graph node is constructed by treating the relation of each support–query pair as a graph node in the preprocess module. After that, in the relation propagation module (RPM), the valuable information of support–query pairs is modelled and propagated to directly generate the relational representations for further prediction. Then, a relation constraint module is introduced to regularise the relational representations and make it consistent with the ground‐truth relations as much as possible. With the guidance of the effective RPM and relation constraint module, the relational representations of the support–query pairs are distinguishable and thus can achieve accurate predictive results. Comprehensive experiments conducted on widely used benchmarks validate the effectiveness of our method compared to state‐of‐the‐art few‐shot classification approaches.

Achieving an automatic trade-off between accuracy and efficiency for a single deep neural network is highly desired in time-sensitive computer vision applications. To achieve anytime prediction, existing methods only embed fixed exits to neural networks and make the predictions with the fixed exits for all the samples (refer to the "latest-all" strategy). However, it is observed that the latest exit within a time budget does not always provide a more accurate prediction than the earlier exits for testing samples of various difficulties, making the "latest-all" strategy a sub-optimal solution. Motivated by this, we propose to improve the anytime prediction accuracy by allowing each sample to adaptively select its own optimal exit within a specific time budget. Specifically, we propose a new Routing Convolutional Network (RCN). For any given time budget, it adaptively selects the optimal layer as exit for a specific testing sample. To learn an optimal policy for sample routing, a Q-network is embedded into the RCN at each exit, considering both potential information gain and time-cost. To further boost the anytime prediction accuracy, the exits and the Q-networks are optimized alternately to mutually boost each other under the cost-sensitive environment. Apart from applying to whole image classification, RCN can also be adapted to dense prediction tasks, e.g., scene parsing, to achieve the pixel-level anytime prediction. Extensive experimental results on CIFAR-10, CIFAR-100, and ImageNet classification benchmarks, and Cityscapes scene parsing benchmark demonstrate the efficacy of the proposed RCN for anytime recognition.

In this work, we aim to address the problem of human interaction recognition in videos by exploring the long-term inter-related dynamics among multiple persons. Recently, Long Short-Term Memory (LSTM) has become a popular choice to model individual dynamic for single-person action recognition due to its ability to capture the temporal motion information in a range. However, most existing LSTM-based methods focus only on capturing the dynamics of human interaction by simply combining all dynamics of individuals or modeling them as a whole. Such methods neglect the inter-related dynamics of how human interactions change over time. To this end, we propose a novel Hierarchical Long Short-Term Concurrent Memory (H-LSTCM) to model the long-term inter-related dynamics among a group of persons for recognizing human interactions. Specifically, we first feed each person's static features into a Single-Person LSTM to model the single-person dynamic. Subsequently, at one time step, the outputs of all Single-Person LSTM units are fed into a novel Concurrent LSTM (Co-LSTM) unit, which mainly consists of multiple sub-memory units, a new cell gate, and a new co-memory cell. In the Co-LSTM unit, each sub-memory unit stores individual motion information, while this Co-LSTM unit selectively integrates and stores inter-related motion information between multiple interacting persons from multiple sub-memory units via the cell gate and co-memory cell, respectively. Extensive experiments on several public datasets validate the effectiveness of the proposed H-LSTCM by comparing against baseline and state-of-the-art methods.

In this paper, we propose an end-to-end deep learning architecture that generates 3D triangular meshes from single color images. Restricted by the nature of prevalent deep learning techniques, the majority of previous works represent 3D shapes in volumes or point clouds. However, it is non-trivial to convert these representations to compact and ready-to-use mesh models. Unlike the existing methods, our network represents 3D shapes in meshes, which are essentially graphs and well suited for graph-based convolutional neural networks. Leveraging perceptual features extracted from an input image, our network produces the correct geometry by progressively deforming an ellipsoid. To make the whole deformation procedure stable, we adopt a coarse-to-fine strategy, and define various mesh/surface related losses to capture properties of various aspects, which benefits producing the visually appealing and physically accurate 3D geometry. In addition, our model by nature can be adapted to objects in specific domains, e.g., human faces, and be easily extended to learn per-vertex properties, e.g., color. Extensive experiments show that our method not only qualitatively produces the mesh model with better details, but also achieves the higher 3D shape estimation accuracy compared against the state-of-the-arts.

In this article, we present a distributed variant of an adaptive stochastic gradient method for training deep neural networks in the parameter-server model. To reduce the communication cost among the workers and server, we incorporate two types of quantization schemes, i.e., gradient quantization and weight quantization, into the proposed distributed Adam. In addition, to reduce the bias introduced by quantization operations, we propose an error-feedback technique to compensate for the quantized gradient. Theoretically, in the stochastic nonconvex setting, we show that the distributed adaptive gradient method with gradient quantization and error feedback converges to the first-order stationary point, and that the distributed adaptive gradient method with weight quantization and error feedback converges to the point related to the quantized level under both the single-worker and multi-worker modes. Last, we apply the proposed distributed adaptive gradient methods to train deep neural networks. Experimental results demonstrate the efficacy of our methods.

Learning representation from relative similarity comparisons, often called ordinal embedding, gains rising attention in recent years. Most of the existing methods are based on semi-definite programming (SDP), which is generally time-consuming and degrades the scalability, especially confronting large-scale data. To overcome this challenge, we propose a stochastic algorithm called SVRG-SBB, which has the following features: i) achieving good scalability via dropping positive semi-definite (PSD) constraints as serving a fast algorithm, i.e., stochastic variance reduced gradient (SVRG) method, and ii) adaptive learning via introducing a new, adaptive step size called the stabilized Barzilai-Borwein (SBB) step size. Theoretically, under some natural assumptions, we show theO(1/T) O(1T) rate of convergence to a stationary point of the proposed algorithm, where T T is the number of total iterations. Under the further Polyak-Łojasiewicz assumption, we can show the global linear convergence (i.e., exponentially fast converging to a global optimum) of the proposed algorithm. Numerous simulations and real-world data experiments are conducted to show the effectiveness of the proposed algorithm by comparing with the state-of-the-art methods, notably, much lower computational cost with good prediction performance.

Bounding box regression is the crucial step in object detection. In existing methods, while ℓn-norm loss is widely adopted for bounding box regression, it is not tailored to the evaluation metric, i.e., Intersection over Union (IoU). Recently, IoU loss and generalized IoU (GIoU) loss have been proposed to benefit the IoU metric, but still suffer from the problems of slow convergence and inaccurate regression. In this paper, we propose a Distance-IoU (DIoU) loss by incorporating the normalized distance between the predicted box and the target box, which converges much faster in training than IoU and GIoU losses. Furthermore, this paper summarizes three geometric factors in bounding box regression, i.e., overlap area, central point distance and aspect ratio, based on which a Complete IoU (CIoU) loss is proposed, thereby leading to faster convergence and better performance. By incorporating DIoU and CIoU losses into state-of-the-art object detection algorithms, e.g., YOLO v3, SSD and Faster R-CNN, we achieve notable performance gains in terms of not only IoU metric but also GIoU metric. Moreover, DIoU can be easily adopted into non-maximum suppression (NMS) to act as the criterion, further boosting performance improvement. The source code and trained models are available at https://github.com/Zzh-tju/DIoU.

We propose to incorporate neural architecture search (NAS) into general-purpose multi-task learning (GP-MTL). Existing NAS methods typically define different search spaces according to different tasks. In order to adapt to different task combinations (i.e., task sets), we disentangle the GP-MTL networks into single-task backbones (optionally encode the task priors), and a hierarchical and layerwise features sharing/fusing scheme across them. This enables us to design a novel and general task-agnostic search space, which inserts cross-task edges (i.e., feature fusion connections) into fixed single-task network backbones. Moreover, we also propose a novel single-shot gradient-based search algorithm that closes the performance gap between the searched architectures and the final evaluation architecture. This is realized with a minimum entropy regularization on the architecture weights during the search phase, which makes the architecture weights converge to near-discrete values and therefore achieves a single model. As a result, our searched model can be directly used for evaluation without (re-)training from scratch. We perform extensive experiments using different single-task backbones on various task sets, demonstrating the promising performance obtained by exploiting the hierarchical and layerwise features, as well as the desirable generalizability to different i) task sets and ii) single-task backbones. The code of our paper is available at https://github.com/bhpfelix/MTLNAS.

Image visual try-on aims at transferring a target clothes image onto a reference person, and has become a hot topic in recent years. Prior arts usually focus on preserving the character of a clothes image (e.g. texture, logo, embroidery) when warping it to arbitrary human pose. However, it remains a big challenge to generate photo-realistic try-on images when large occlusions and human poses are presented in the reference person. To address this issue, we propose a novel visual try-on network, namely Adaptive Content Generating and Preserving Network (ACGPN). In particular, ACGPN first predicts semantic layout of the reference image that will be changed after try-on (e.g.long sleeve shirt→arm, arm→jacket), and then determines whether its image content needs to be generated or preserved according to the predicted semantic layout, leading to photo-realistic try-on and rich clothes details. ACGPN generally involves three major modules. First, a semantic layout generation module utilizes semantic segmentation of the reference image to progressively predict the desired semantic layout after try-on. Second, a clothes warping module warps clothes image according to the generated semantic layout, where a second-order difference constraint is introduced to stabilize the warping process during training.Third, an inpainting module for content fusion integrates all information (e.g. reference image, semantic layout, warped clothes) to adaptively produce each semantic part of human body. In comparison to the state-of-the-art methods, ACGPN can generate photo-realistic images with much better perceptual quality and richer fine-details.

Existing data-dependent hashing methods usually learn hash functions from pairwise or triplet data relationships, which only capture the data similarity locally, and often suffer from low learning efficiency and low collision rate. In this work, we propose a new \emph{global} similarity metric, termed as \emph{central similarity}, with which the hash codes of similar data pairs are encouraged to approach a common center and those for dissimilar pairs to converge to different centers, to improve hash learning efficiency and retrieval accuracy. We principally formulate the computation of the proposed central similarity metric by introducing a new concept, i.e., \emph{hash center} that refers to a set of data points scattered in the Hamming space with a sufficient mutual distance between each other. We then provide an efficient method to construct well separated hash centers by leveraging the Hadamard matrix and Bernoulli distributions. Finally, we propose the Central Similarity Quantization (CSQ) that optimizes the central similarity between data points w.r.t.\ their hash centers instead of optimizing the local similarity. CSQ is generic and applicable to both image and video hashing scenarios. Extensive experiments on large-scale image and video retrieval tasks demonstrate that CSQ can generate cohesive hash codes for similar data pairs and dispersed hash codes for dissimilar pairs, achieving a noticeable boost in retrieval performance, i.e. 3\%-20\% in mAP over the previous state-of-the-arts.

Existing deep learning methods for image deblurring typically train models using pairs of sharp images and their blurred counterparts. However, synthetically blurring images does not necessarily model the blurring process in real-world scenarios with sufficient accuracy. To address this problem, we propose a new method which combines two GAN models, i.e., a learning-to-Blur GAN (BGAN) and learning-to-DeBlur GAN (DBGAN), in order to learn a better model for image deblurring by primarily learning how to blur images. The first model, BGAN, learns how to blur sharp images with unpaired sharp and blurry image sets, and then guides the second model, DBGAN, to learn how to correctly deblur such images. In order to reduce the discrepancy between real blur and synthesized blur, a relativistic blur loss is leveraged. As an additional contribution, this paper also introduces a Real-World Blurred Image (RWBI) dataset including diverse blurry images. Our experiments show that the proposed method achieves consistently superior quantitative performance as well as higher perceptual quality on both the newly proposed dataset and the public GOPRO dataset.

In this paper, we study distributed algorithms for large-scale AUC maximization with a deep neural network as a predictive model. Although distributed learning techniques have been investigated extensively in deep learning, they are not directly applicable to stochastic AUC maximization with deep neural networks due to its striking differences from standard loss minimization problems (e.g., cross-entropy). Towards addressing this challenge, we propose and analyze a communication-efficient distributed optimization algorithm based on a \emph{non-convex concave} reformulation of the AUC maximization, in which the communication of both the primal variable and the dual variable between each worker and the parameter server only occurs after multiple steps of gradient-based updates in each worker. Compared with the naive parallel version of an existing algorithm that computes stochastic gradients at individual machines and averages them for updating the model parameter, our algorithm requires a much less number of communication rounds and still achieves linear speedup in theory. To the best of our knowledge, this is the \textbf{first} work that solves the \emph{non-convex concave min-max} problem for AUC maximization with deep neural networks in a communication-efficient distributed manner while still maintaining the linear speedup property in theory. Our experiments on several benchmark datasets show the effectiveness of our algorithm and also confirm our theory.

Few-shot learning, aiming to learn novel concepts from few labeled examples, is an interesting and very challenging problem with many practical advantages. To accomplish this task, one should concentrate on revealing the accurate relations of the support-query pairs. We propose a transductive relation-propagation graph neural network (TRPN) to explicitly model and propagate such relations across support-query pairs. Our TRPN treats the relation of each support-query pair as a graph node, named relational node, and resorts to the known relations between support samples, including both intra-class commonality and inter-class uniqueness, to guide the relation propagation in the graph, generating the discriminative relation embeddings for support-query pairs. A pseudo relational node is further introduced to propagate the query characteristics, and a fast, yet effective transductive learning strategy is devised to fully exploit the relation information among different queries. To the best of our knowledge, this is the first work that explicitly takes the relations of support-query pairs into consideration in few-shot learning, which might offer a new way to solve the few-shot learning problem. Extensive experiments conducted on several benchmark datasets demonstrate that our method can significantly outperform a variety of state-of-the-art few-shot learning methods.

Few-shot learning aims to learn a model that can be readily adapted to new unseen classes (concepts) by accessing one or few examples. Despite the successful progress, most of the few-shot learning approaches, concentrating on either global or local characteristics of examples, still suffer from weak generalization abilities. Inspired by the inverted pyramid theory, to address this problem, we propose an inverted pyramid network (IPN) that intimates the human's coarse-to-fine cognition paradigm. The proposed IPN consists of two consecutive stages, namely global stage and local stage. At the global stage, a class-sensitive contextual memory network (CCMNet) is introduced to learn discriminative support-query relation embeddings and predict the query-to-class similarity based on the contextual memory. Then at the local stage, a fine-grained calibration is further appended to complement the coarse relation embeddings, targeting more precise query-to-class similarity evaluation. To the best of our knowledge, IPN is the first work that simultaneously integrates both global and local characteristics in few-shot learning, approximately imitating the human cognition mechanism. Our extensive experiments on multiple benchmark datasets demonstrate the superiority of IPN, compared to a number of state-of-the-art approaches.

The vulnerability of deep neural networks has gained a great upsurge of research attention, which engages well-designed examples through adding little perturbations to fool a well-performed network. Meanwhile, a progress has been made in leveraging adversarial examples to boost the robustness of deep cross-modal networks. However, for cross-modal learning, both the causes of adversarial examples and their latent advantages in learning cross-modal correlations are under-explored. In this paper, we propose novel Disentangled Adversarial examples for Cross-Modal learning, dubbed DACM. Specifically, we first divide cross-modal data into two aspects, namely modality-related component and modality-unrelated counterpart, and then learn to improve the reliability of network using the modality-related component. To achieve this goal, we apply the generation of adversarial perturbations to strengthen cross-modal correlations, wherein the modality-related component is acquired through gradually detaching the modality-unrelated component. Finally, the proposed DACM is employed to create modality-related examples towards the application of cross-modal hashing retrieval. Extensive experiments carried out on two cross-modal benchmarks show that the adversarial examples learned by DACM are efficient at fooling a target deep cross-modal hashing network. On the other hand, training this target model by merely leveraging our created modality-related examples in turn significantly promotes the robustness of this model itself.

Epoch gradient descent method (a.k.a. Epoch-GD) proposed by (Hazan and Kale, 2011) was deemeda breakthrough for stochastic strongly convex minimization, which achieves theoptimal convergence rate of O(1/T) with T iterative updates for the objective gap. However, its extension to solving stochastic min-max problems with strong convexity and strong concavity still remains open, and it is still unclear whethera fast rate ofO(1/T)for theduality gapis achievable for stochastic min-max optimization under strong convexity and strong concavity. Although some re-cent studies have proposed stochastic algorithms with fast convergence rates formin-max problems, they require additional assumptions about the problem, e.g.,smoothness, bi-linear structure, etc. In this paper, we bridge this gap by providinga sharp analysis of epoch-wise stochastic gradient descent ascent method (referredto as Epoch-GDA) for solving strongly convex strongly concave (SCSC) min-maxproblems, without imposing any additional assumption about smoothness or the function’s structure. To the best of our knowledge, our result is the first one that shows Epoch-GDA can achieve the optimal rate ofO(1/T)for the duality gapof general SCSC min-max problems. We emphasize that such generalization of Epoch-GD for strongly convex minimization problems to Epoch-GDA for SCSC min-max problems is non-trivial and requires novel technical analysis. Moreover, we notice that the key lemma can also be used for proving the convergence of Epoch-GDA for weakly-convex strongly-concave min-max problems, leading to a nearly optimal complexity without resorting to smoothness or other structural conditions.

Deep clustering integrates embedding and clustering together to obtain the optimal nonlinear embedding space, which is more effective in real-world scenarios compared with conventional clustering methods. However, the robustness of the clustering network is prone to being attenuated especially when it encounters an adversarial attack. A small perturbation in the embedding space will lead to diverse clustering results since the labels are absent. In this paper, we propose a robust deep clustering method based on adversarial learning. Specifically, we first attempt to define adversarial samples in the embedding space for the clustering network. Meanwhile, we devise an adversarial attack strategy to explore samples that easily fool the clustering layers but do not impact the performance of the deep embedding. We then provide a simple yet efficient defense algorithm to improve the robustness of the clustering network. Experimental results on two popular datasets show that the proposed adversarial learning method can significantly enhance the robustness and further improve the overall clustering performance. Particularly, the proposed method is generally applicable to multiple existing clustering frameworks to boost their robustness. The source code is available at https://github.com/xdxuyang/ALRDC.

MOBA games, e.g., Honor of Kings, League of Legends, and Dota 2, pose grand challenges to AI systems such as multi-agent, enormous state-action space, complex action control, etc. Developing AI for playing MOBA games has raised much attention accordingly. However, existing work falls short in handling the raw game complexity caused by the explosion of agent combinations, i.e., lineups, when expanding the hero pool in case that OpenAI's Dota AI limits the play to a pool of only 17 heroes. As a result, full MOBA games without restrictions are far from being mastered by any existing AI system. In this paper, we propose a MOBA AI learning paradigm that methodologically enables playing full MOBA games with deep reinforcement learning. Specifically, we develop a combination of novel and existing learning techniques, including off-policy adaption, multi-head value estimation, curriculum self-play learning, policy distillation, and Monte-Carlo tree-search, in training and playing a large pool of heroes, meanwhile addressing the scalability issue skillfully. Tested on Honor of Kings, a popular MOBA game, we show how to build superhuman AI agents that can defeat top esports players. The superiority of our AI is demonstrated by the first large-scale performance test of MOBA AI agent in the literature.

Deep metric learning plays a key role in various machine learning tasks. Most of the previous works have been confined to sampling from a mini-batch, which cannot precisely characterize the global geometry of the embedding space. Although researchers have developed proxy- and classification-based methods to tackle the sampling issue, those methods inevitably incur a redundant computational cost. In this paper, we propose a novel Proxy-based deep Graph Metric Learning (ProxyGML) approach from the perspective of graph classification, which uses fewer proxies yet achieves better comprehensive performance. Specifically, multiple global proxies are leveraged to collectively approximate the original data points for each class. To efficiently capture local neighbor relationships, a small number of such proxies are adaptively selected to construct similarity subgraphs between these proxies and each data point. Further, we design a novel reverse label propagation algorithm, by which the neighbor relationships are adjusted according to ground-truth labels, so that a discriminative metric space can be learned during the process of subgraph classification. Extensive experiments carried out on widely-used CUB-200-2011, Cars196, and Stanford Online Products datasets demonstrate the superiority of the proposed ProxyGML over the state-of-the-art methods in terms of both effectiveness and efficiency. The source code is publicly available at \url{https://github.com/YuehuaZhu/ProxyGML}.

In this paper, we consider the Tensor Robust Principal Component Analysis (TRPCA) problem, which aims to exactly recover the low-rank and sparse components from their sum. Our model is based on the recently proposed tensor-tensor product (or t-product) [14]. Induced by the t-product, we first rigorously deduce the tensor spectral norm, tensor nuclear norm, and tensor average rank, and show that the tensor nuclear norm is the convex envelope of the tensor average rank within the unit ball of the tensor spectral norm. These definitions, their relationships and properties are consistent with matrix cases. Equipped with the new tensor nuclear norm, we then solve the TRPCA problem by solving a convex program and provide the theoretical guarantee for the exact recovery. Our TRPCA model and recovery guarantee include matrix RPCA as a special case. Numerical experiments verify our results, and the applications to image recovery and background modeling problems demonstrate the effectiveness of our method.

We study active object tracking, where a tracker takes visual observations (i.e., frame sequences) as input and produces the corresponding camera control signals as output (e.g., move forward, turn left, etc.). Conventional methods tackle tracking and camera control tasks separately, and the resulting system is difficult to tune jointly. These methods also require significant human efforts for image labeling and expensive trial-and-error system tuning in the real world. To address these issues, we propose, in this paper, an end-to-end solution via deep reinforcement learning. A ConvNet-LSTM function approximator is adopted for the direct frame-to-action prediction. We further propose an environment augmentation technique and a customized reward function, which are crucial for successful training. The tracker trained in simulators (ViZDoom and Unreal Engine) demonstrates good generalization behaviors in the case of unseen object moving paths, unseen object appearances, unseen backgrounds, and distracting objects. The system is robust and can restore tracking after occasional lost of the target being tracked. We also find that the tracking ability, obtained solely from simulators, can potentially transfer to real-world scenarios. We demonstrate successful examples of such transfer, via experiments over the VOT dataset and the deployment of a real-world robot using the proposed active tracker trained in simulation.

In this paper, the problem of describing visual contents of a video sequence with natural language is addressed. Unlike previous video captioning work mainly exploiting the cues of video contents to make a language description, we propose a reconstruction network (RecNet) in a novel encoder-decoder-reconstructor architecture, which leverages both forward (video to sentence) and backward (sentence to video) flows for video captioning. Specifically, the encoder-decoder component makes use of the forward flow to produce a sentence based on the encoded video semantic features. Two types of reconstructors are subsequently proposed to employ the backward flow and reproduce the video features from local and global perspectives, respectively, capitalizing on the hidden state sequence generated by the decoder. Moreover, in order to make a comprehensive reconstruction of the video features, we propose to fuse the two types of reconstructors together. The generation loss yielded by the encoder-decoder component and the reconstruction loss introduced by the reconstructor are jointly cast into training the proposed RecNet in an end-to-end fashion. Furthermore, the RecNet is fine-tuned by CIDEr optimization via reinforcement learning, which significantly boosts the captioning performance. Experimental results on benchmark datasets demonstrate that the proposed reconstructor can boost the performance of video captioning consistently.

In this paper, we propose a novel multimodal matching model for the image and sentence based on their multiple representations. Each representation of the image or sentence undergoes an independent neural network, consisting of multiple layers of nonlinear mappings to yield the corresponding embedding. Besides exploiting the image and sentence relationship based on their embeddings, we propose one novel loss to further exploit the relationship within each single modality, namely, image and sentence based on the yielded multiple embeddings, which is used to train the neural networks simultaneously. The experimental results demonstrate that multiple representations can help to capture the image contents and the sentence semantic meaning more precisely, thus making comprehensive exploitations of the complicated image and sentence matching relationship. More concretely, the proposed matching model significantly outperforms the state-of-the-art approaches in bidirectional image-sentence retrieval on the Flickr8K, Flickr30K, and Microsoft COCO datasets.

Recently, graph-based hashing that learns similarity-preserving binary codes via an affinity graph has been extensively studied for large-scale image retrieval. However, most graph-based hashing methods resort to intractable binary quadratic programs, making them unscalable to massive data. In this paper, we propose a novel graph convolutional network-based hashing framework, dubbed GCNH, which directly carries out spectral convolution operations on both an image set and an affinity graph built over the set, naturally yielding similarity-preserving binary embedding. GCNH fundamentally differs from conventional graph hashing methods which adopt an affinity graph as the only learning guidance in an objective function to pursue the binary embedding. As the core ingredient of GCNH, we introduce an intuitive asymmetric graph convolutional (AGC) layer to simultaneously convolve the anchor graph, input data, and convolutional filters. By virtue of the AGC layer, GCNH well addresses the issues of scalability and out-of-sample extension when leveraging affinity graphs for hashing. As a use case of our GCNH, we particularly study the semisupervised hashing scenario in this paper. Comprehensive image retrieval evaluations on the CIFAR-10, NUS-WIDE, and ImageNet datasets demonstrate the consistent advantages of GCNH over the state-of-the-art methods given limited labeled data.

Compositing is one of the most important editing operations for images and videos. The process of improving the realism of composite results is often called harmonization. Previous approaches for harmonization mainly focus on images. In this paper, we take one step further to attack the problem of video harmonization. Specifically, we train a convolutional neural network in an adversarial way, exploiting a pixel-wise disharmony discriminator to achieve more realistic harmonized results and introducing a temporal loss to increase temporal consistency between consecutive harmonized frames. Thanks to the pixel-wise disharmony discriminator, we are also able to relieve the need of input foreground masks. Since existing video datasets which have ground-truth foreground masks and optical flows are not sufficiently large, we propose a simple yet efficient method to build up a synthetic dataset supporting supervised training of the proposed adversarial network. The experiments show that training on our synthetic dataset generalizes well to the real-world composite dataset. In addition, our method successfully incorporates temporal consistency during training and achieves more harmonious visual results than previous methods.

Recently, hashing has attracted much attention in visual information retrieval due to its low storage cost and fast query speed. The goal of hashing is to map original high-dimensional data into a low-dimensional binary-code space where the similar data points are assigned similar hash codes and dissimilar points are far away from each other. Existing unsupervised hashing methods mainly focus on recovering the pairwise similarity of the original data in hash space, but do not take specific measures to make the generated binary codes to be discriminative. To address this problem, this paper proposes a novel unsupervised hashing method, named “Discriminative Unsupervised Graph Hashing” (DUGH), which takes both similarity and dissimilarity of original data into consideration to learn discriminative binary codes. In particular, a probabilistic model is utilized to learn the encoding of original data in low-dimensional space, which models the original neighbor structure through both positive and negative edges in the KNN graph and then maximizes the likelihood of observing these edges. To efficiently and accurately measure the neighbor structure for largescale datasets, we propose an effective KNN graph construction algorithm based on the random projection tree and neighbor exploring techniques. The experimental results on one synthetic dataset and four typical real-world image datasets demonstrate that the proposed method significantly outperforms the state-of-the-art unsupervised hashing methods.

In this work, we study the robustness of a CNN+RNN based image captioning system being subjected to adversarial noises. We propose to fool an image captioning system to generate some targeted partial captions for an image polluted by adversarial noises, even the targeted captions are totally irrelevant to the image content. A partial caption indicates that the words at some locations in this caption are observed, while words at other locations are not restricted. It is the first work to study exact adversarial attacks of targeted partial captions. Due to the sequential dependencies among words in a caption, we formulate the generation of adversarial noises for targeted partial captions as a structured output learning problem with latent variables. Both the generalized expectation maximization algorithm and structural SVMs with latent variables are then adopted to optimize the problem. The proposed methods generate very successful attacks to three popular CNN+RNN based image captioning models. Furthermore, the proposed attack methods are used to understand the inner mechanism of image captioning systems, providing the guidance to further improve automatic image captioning systems towards human captioning.

There has been an increasing research interest in age-invariant face recognition. However, matching faces with big age gaps remains a challenging problem, primarily due to the significant discrepancy of face appearance caused by aging. To reduce such discrepancy, in this paper we present a novel algorithm to remove age-related components from features mixed with both identity and age information. Specifically, we factorize a mixed face feature into two uncorrelated components: identity-dependent component and age-dependent component, where the identity-dependent component contains information that is useful for face recognition. To implement this idea, we propose the Decorrelated Adversarial Learning (DAL) algorithm, where a Canonical Mapping Module (CMM) is introduced to find maximum correlation of the paired features generated by the backbone network, while the backbone network and the factorization module are trained to generate features reducing the correlation. Thus, the proposed model learns the decomposed features of age and identity whose correlation is significantly reduced. Simultaneously, the identity-dependent feature and the age-dependent feature are supervised by ID and age preserving signals respectively to ensure they contain the correct information. Extensive experiments have been conducted on the popular public-domain face aging datasets (FG-NET, MORPH Album 2, and CACD-VS) to demonstrate the effectiveness of the proposed approach.

Humans can robustly learn novel visual concepts even when images undergo various deformations and loose certain information. Mimicking the same behavior and synthesizing deformed instances of new concepts may help visual recognition systems perform better one-shot learning, i.e., learning concepts from one or few examples. Our key insight is that, while the deformed images may not be visually realistic, they still maintain critical semantic information and contribute significantly to formulating classifier decision boundaries. Inspired by the recent progress of meta-learning, we combine a meta-learner with an image deformation sub-network that produces additional training examples, and optimize both models in an end-to-end manner. The deformation sub-network learns to deform images by fusing a pair of images --- a probe image that keeps the visual content and a gallery image that diversifies the deformations. We demonstrate results on the widely used one-shot learning benchmarks (miniImageNet and ImageNet 1K Challenge datasets), which significantly outperform state-of-the-art approaches.

Face recognition has obtained remarkable progress in recent years due to the great improvement of deep convolutional neural networks (CNNs). However, deep CNNs are vulnerable to adversarial examples, which can cause fateful consequences in real-world face recognition applications with security-sensitive purposes. Adversarial attacks are widely studied as they can identify the vulnerability of the models before they are deployed. In this paper, we evaluate the robustness of state-of-the-art face recognition models in the decision-based black-box attack setting, where the attackers have no access to the model parameters and gradients, but can only acquire hard-label predictions by sending queries to the target model. This attack setting is more practical in real-world face recognition systems. To improve the efficiency of previous methods, we propose an evolutionary attack algorithm, which can model the local geometry of the search directions and reduce the dimension of the search space. Extensive experiments demonstrate the effectiveness of the proposed method that induces a minimum perturbation to an input face image with fewer queries. We also apply the proposed method to attack a real-world face recognition system successfully.

The need for efficiently finding the video content a user wants is increasing because of the erupting of user-generated videos on the Web. Existing keyword-based or content-based video retrieval methods usually determine what occurs in a video but not when and where. In this paper, we make an answer to the question of when and where by formulating a new task, namely spatio-temporal video re-localization. Specifically, given a query video and a reference video, spatio-temporal video re-localization aims to localize tubelets in the reference video such that the tubelets semantically correspond to the query. To accurately localize the desired tubelets in the reference video, we propose a novel warp LSTM network, which propagates the spatio-temporal information for a long period and thereby captures the corresponding long-term dependencies. Another issue for spatio-temporal video re-localization is the lack of properly labeled video datasets. Therefore, we reorganize the videos in the AVA dataset to form a new dataset for spatio-temporal video re-localization research. Extensive experimental results show that the proposed model achieves superior performances over the designed baselines on the spatio-temporal video re-localization task.

Deep neural networks have achieved great successes on the image captioning task. However, most of the existing models depend heavily on paired image-sentence datasets, which are very expensive to acquire. In this paper, we make the first attempt to train an image captioning model in an unsupervised manner. Instead of relying on manually labeled image-sentence pairs, our proposed model merely requires an image set, a sentence corpus, and an existing visual concept detector. The sentence corpus is used to teach the captioning model how to generate plausible sentences. Meanwhile, the knowledge in the visual concept detector is distilled into the captioning model to guide the model to recognize the visual concepts in an image. In order to further encourage the generated captions to be semantically consistent with the image, the image and caption are projected into a common latent space so that they can reconstruct each other. Given that the existing sentence corpora are mainly designed for linguistic research and are thus with little reference to image contents, we crawl a large-scale image description corpus of two million natural sentences to facilitate the unsupervised image captioning scenario. Experimental results show that our proposed model is able to produce quite promising results without any caption annotations.

In this paper, we propose a novel Convolutional Neural Network (CNN) structure for general-purpose multi-task learning (MTL), which enables automatic feature fusing at every layer from different tasks. This is in contrast with the most widely used MTL CNN structures which empirically or heuristically share features on some specific layers (e.g., share all the features except the last convolutional layer). The proposed layerwise feature fusing scheme is formulated by combining existing CNN components in a novel way, with clear mathematical interpretability as discriminative dimensionality reduction, which is referred to as Neural Discriminative Dimensionality Reduction (NDDR). Specifically, we first concatenate features with the same spatial resolution from different tasks according to their channel dimension. Then, we show that the discriminative dimensionality reduction can be fulfilled by 1×1 Convolution, Batch Normalization, and Weight Decay in one CNN. The use of existing CNN components ensures the end-to-end training and the extensibility of the proposed NDDR layer to various state-of-the-art CNN architectures in a "plug-and-play" manner. The detailed ablation analysis shows that the proposed NDDR layer is easy to train and also robust to different hyperparameters. Experiments on different task sets with various base network architectures demonstrate the promising performance and desirable generalizability of our proposed method. The code of our paper is available at https://github.com/ethanygao/NDDR-CNN.

This work studies the model compression for deep convolutional neural networks (CNNs) via filter pruning. The workflow of a traditional pruning consists of three sequential stages: pre-training the original model, selecting the pre-trained filters via ranking according to a manually designed criterion (e.g., the norm of filters), and learning the remained filters via fine-tuning. Most existing works follow this pipeline and focus on designing different ranking criteria for filter selection. However, it is difficult to control the performance due to the separation of filter selection and filter learning. In this work, we propose to conduct filter selection and filter learning simultaneously, in a unified model. To this end, we define a factorized convolutional filter (FCF), consisting of a standard real-valued convolutional filter and a binary scalar, as well as a dot-product operator between them. We train a CNN model with factorized convolutional filters (CNN-FCF) by updating the standard filter using back-propagation, while updating the binary scalar using the alternating direction method of multipliers (ADMM) based optimization method. With this trained CNN-FCF model, we only keep the standard filters corresponding to the 1-valued scalars, while all other filters and all binary scalars are discarded, to obtain a compact CNN model. Extensive experiments on CIFAR-10 and ImageNet demonstrate the superiority of the proposed method over state-of-the-art filter pruning methods.

Temporal action proposal generation is an important task, aiming to localize the video segments containing human actions in an untrimmed video. In this paper, we propose a multi-granularity generator (MGG) to perform the temporal action proposal from different granularity perspectives, relying on the video visual features equipped with the position embedding information. First, we propose to use a bilinear matching model to exploit the rich local information within the video sequence. Afterwards, two components, namely segment proposal producer (SPP) and frame actionness producer (FAP), are combined to perform the task of temporal action proposal at two distinct granularities. SPP considers the whole video in the form of feature pyramid and generates segment proposals from one coarse perspective, while FAP carries out a finer actionness evaluation for each video frame. Our proposed MGG can be trained in an end-to-end fashion. Through temporally adjusting the segment proposals with fine-grained information based on frame actionness, MGG achieves the superior performance over state-of-the-art methods on the public THUMOS-14 and ActivityNet-1.3 datasets. Moreover, we employ existing action classifiers to perform the classification of the proposals generated by MGG, leading to significant improvements compared against the competing methods for the video detection task.

We propose to compose dynamic tree structures that place the objects in an image into a visual context, helping visual reasoning tasks such as scene graph generation and visual Q&A. Our visual context tree model, dubbed VCTree, has two key advantages over existing structured object representations including chains and fully-connected graphs: 1) The efficient and expressive binary tree encodes the inherent parallel/hierarchical relationships among objects, e.g., ``clothes'' and ``pants'' are usually co-occur and belong to ``person''; 2) the dynamic structure varies from image to image and task to task, allowing more content-/task-specific message passing among objects. To construct a VCTree, we design a score function that calculates the task-dependent validity between each object pair, and the tree is the binary version of the maximum spanning tree from the score matrix. Then, visual contexts are encoded by bidirectional TreeLSTM and decoded by task-specific models. We develop a hybrid learning procedure which integrates end-task supervised learning and the tree structure reinforcement learning, where the former's evaluation result serves as a self-critic for the latter's structure exploration. Experimental results on two benchmarks, which require reasoning over contexts: Visual Genome for scene graph generation and VQA2.0 for visual Q&A, show that VCTree outperforms state-of-the-art results while discovering interpretable visual context structures.

Crowd counting is a challenging task due to factors such as large variations in crowdedness and severe occlusions. Although recent deep learning based counting algorithms have achieved a great progress, the correlation knowledge among samples and the semantic prior have not yet been fully exploited. In this paper, a residual regression framework is proposed for crowd counting utilizing the correlation information among samples. By incorporating such information into our network, we discover that more intrinsic characteristics can be learned by the network which thus generalizes better to unseen scenarios. Besides, we show how to effectively leverage the semantic prior to improve the performance of crowd counting. We also observe that the adversarial loss can be used to improve the quality of predicted density maps, thus leading to an improvement in crowd counting. Experiments on public datasets demonstrate the effectiveness and generalization ability of the proposed method.

We address the problem of video representation learning without human-annotated labels. While previous efforts address the problem by designing novel self-supervised tasks using video data, the learned features are merely on a frame-by-frame basis, which are not applicable to many video analytic tasks where spatio-temporal features are prevailing. In this paper we propose a novel self-supervised approach to learn spatio-temporal features for video representation. Inspired by the success of two-stream approaches in video classification, we propose to learn visual features by regressing both motion and appearance statistics along spatial and temporal dimensions, given only the input video data. Specifically, we extract statistical concepts (fast-motion region and the corresponding dominant direction, spatio-temporal color diversity, dominant color, etc.) from simple patterns in both spatial and temporal domains. Unlike prior puzzles that are even hard for humans to solve, the proposed approach is consistent with human inherent visual habits and therefore easy to answer. We conduct extensive experiments with C3D to validate the effectiveness of our proposed approach. The experiments show that our approach can significantly improve the performance of C3D when applied to video classification tasks. Code is available at https://github.com/laura-wang/video_repres_mas.

We propose an unsupervised visual tracking method in this paper. Different from existing approaches using extensive annotated data for supervised learning, our CNN model is trained on large-scale unlabeled videos in an unsupervised manner. Our motivation is that a robust tracker should be effective in both the forward and backward predictions (i.e., the tracker can forward localize the target object in successive frames and backtrace to its initial position in the first frame). We build our framework on a Siamese correlation filter network, which is trained using unlabeled raw videos. Meanwhile, we propose a multiple-frame validation method and a cost-sensitive loss to facilitate unsupervised learning. Without bells and whistles, the proposed unsupervised tracker achieves the baseline accuracy of fully supervised trackers, which require complete and accurate labels during training. Furthermore, unsupervised framework exhibits a potential in leveraging unlabeled or weakly labeled data to further improve the tracking accuracy.

We address the problem of recovering the 3D geometry of a human face from a set of facial images in multiple views. While recent studies have shown impressive progress in 3D Morphable Model (3DMM) based facial reconstruction, the settings are mostly restricted to a single view. There is an inherent drawback in the single-view setting: the lack of reliable 3D constraints can cause unresolvable ambiguities. We in this paper explore 3DMM-based shape recovery in a different setting, where a set of multi-view facial images are given as input. A novel approach is proposed to regress 3DMM parameters from multi-view inputs with an end-to-end trainable Convolutional Neural Network (CNN). Multi-view geometric constraints are incorporated into the network by establishing dense correspondences between different views leveraging a novel self-supervised view alignment loss. The main ingredient of the view alignment loss is a differentiable dense optical flow estimator that can backpropagate the alignment errors between an input view and a synthetic rendering from another input view, which is projected to the target view through the 3D shape to be inferred. Through minimizing the view alignment loss, better 3D shapes can be recovered such that the synthetic projections from one view to another can better align with the observed image. Extensive experiments demonstrate the superiority of the proposed method over other 3DMM methods.

We address the problem of semi-supervised video object segmentation (VOS), where the masks of objects of interests are given in the first frame of an input video. To deal with challenging cases where objects are occluded or missing, previous work relies on greedy data association strategies that make decisions for each frame individually. In this paper, we propose a novel approach to defer the decision making for a target object in each frame, until a global view can be established with the entire video being taken into consideration. Our approach is in the same spirit as Multiple Hypotheses Tracking (MHT) methods, making several critical adaptations for the VOS problem. We employ the bounding box (bbox) hypothesis for tracking tree formation, and the multiple hypotheses are spawned by propagating the preceding bbox into the detected bbox proposals within a gated region starting from the initial object mask in the first frame. The gated region is determined by a gating scheme which takes into account a more comprehensive motion model rather than the simple Kalman filtering model in traditional MHT. To further design more customized algorithms tailored for VOS, we develop a novel mask propagation score instead of the appearance similarity score that could be brittle due to large deformations. The mask propagation score, together with the motion score, determines the affinity between the hypotheses during tree pruning. Finally, a novel mask merging strategy is employed to handle mask conflicts between objects. Extensive experiments on challenging datasets demonstrate the effectiveness of the proposed method, especially in the case of object missing.

The clustering methods have recently absorbed even-increasing attention in learning and vision. Deep clustering combines embedding and clustering together to obtain optimal embedding subspace for clustering, which can be more effective compared with conventional clustering methods. In this paper, we propose a joint learning framework for discriminative embedding and spectral clustering. We first devise a dual autoencoder network, which enforces the reconstruction constraint for the latent representations and their noisy versions, to embed the inputs into a latent space for clustering. As such the learned latent representations can be more robust to noise. Then the mutual information estimation is utilized to provide more discriminative information from the inputs. Furthermore, a deep spectral clustering method is applied to embed the latent representations into the eigenspace and subsequently clusters them, which can fully exploit the relationship between inputs to achieve optimal clustering results. Experimental results on benchmark datasets show that our method can significantly outperform state-of-the-art clustering approaches.

Face anti-spoofing is an important task in full-stack face applications including face detection, verification, and recognition. Previous approaches build models on datasets which do not simulate the real-world data well (e.g., small scale, insignificant variance, etc.). Existing models may rely on auxiliary information, which prevents these anti-spoofing solutions from generalizing well in practice. In this paper, we present a data collection solution along with a data synthesis technique to simulate digital medium-based face spoofing attacks, which can easily help us obtain a large amount of training data well reflecting the real-world scenarios. Through exploiting a novel Spatio-Temporal Anti-Spoof Network (STASN), we are able to push the performance on public face anti-spoofing datasets over state-of-the-art methods by a large margin. Since the proposed model can automatically attend to discriminative regions, it makes analyzing the behaviors of the network possible.We conduct extensive experiments and show that the proposed model can distinguish spoof faces by extracting features from a variety of regions to seek out subtle evidences such as borders, moire patterns, reflection artifacts, etc.

Due to storage and search efficiency, hashing has become significantly prevalent for nearest neighbor search. Particularly, deep hashing methods have greatly improved the search performance, typically under supervised scenarios. In contrast, unsupervised deep hashing models can hardly achieve satisfactory performance due to the lack of supervisory similarity signals. To address this problem, in this paper, we propose a new deep unsupervised hashing model, called DistilHash, which can learn a distilled data set, where data pairs have confident similarity signals. Specifically, we investigate the relationship between the initial but noisy similarity signals learned from local structures and the semantic similarity labels assigned by the optimal Bayesian classifier. We show that, under a mild assumption, some data pairs, of which labels are consistent with those assigned by the optimal Bayesian classifier, can be potentially distilled. With this understanding, we design a simple but effective method to distill data pairs automatically and further adopt a Bayesian learning framework to learn hashing functions from the distilled data set. Extensive experimental results on three widely used benchmark datasets demonstrate that our method achieves state-of-the-art search performance.

Gait recognition is an important biometric method popularly used in video surveillance, where the task is to identify people at a distance by their walking patterns from video sequences. Most of the current successful approaches for gait recognition either use a pair of gait images to form a cross-gait representation or rely on a single gait image for unique-gait representation. These two types of representations emperically complement one another. In this paper, we propose a new Joint Unique-gait and Cross-gait Network (JUCNet), to combine the advantages of unique-gait representation with that of cross-gait representation, leading to an significantly improved performance. Another key contribution of this paper is a novel quintuplet loss function, which simultaneously increases the inter-class differences by pushing representations extracted from different subjects apart and decreases the intra-class variations by pulling representations extracted from the same subject together. Experiments show that our method achieves the state-of-the-art performance tested on standard benchmark datasets, demonstrating its superiority over existing methods.

Facial action unit (AU) intensity is an index to characterize human expressions. Accurate AU intensity estimation depends on three major elements: image representation, intensity estimator, and supervisory information. Most existing methods learn intensity estimator with fixed image representation, and rely on the availability of fully annotated supervisory information. In this paper, a novel general framework for AU intensity estimation is presented, which differs from traditional estimation methods in two aspects. First, rather than keeping image representation fixed, it simultaneously learns representation and intensity estimator to achieve an optimal solution. Second, it allows incorporating weak supervisory training signal from human knowledge (e.g. feature smoothness, label smoothness, label ranking, and positive label), which makes our model trainable even fully annotated information is not available. More specifically, human knowledge is represented as either soft or hard constraints which are encoded as regularization terms or equality/inequality constraints, respectively. On top of our novel framework, we additionally propose an efficient algorithm for optimization based on Alternating Direction Method of Multipliers (ADMM). Evaluations on two benchmark databases show that our method outperforms competing methods under different ratios of AU intensity annotations, especially for small ratios.

Adam and RMSProp are two of the most influential adaptive stochastic algorithms for training deep neural networks, which have been pointed out to be divergent even in the convex setting via a few simple counterexamples. Many attempts, such as decreasing an adaptive learning rate, adopting a big batch size, incorporating a temporal decorrelation technique, seeking an analogous surrogate, etc., have been tried to promote Adam/RMSProp-type algorithms to converge. In contrast with existing approaches, we introduce an alternative easy-to-check sufficient condition, which merely depends on the parameters of the base learning rate and combinations of historical second-order moments, to guarantee the global convergence of generic Adam/RMSProp for solving large-scale non-convex stochastic optimization. Moreover, we show that the convergences of several variants of Adam, such as AdamNC, AdaEMA, etc., can be directly implied via the proposed sufficient condition in the non-convex setting. In addition, we illustrate that Adam is essentially a specifically weighted AdaGrad with exponential moving average momentum, which provides a novel perspective for understanding Adam and RMSProp. This observation coupled with this sufficient condition gives much deeper interpretations on their divergences. At last, we validate the sufficient condition by applying Adam and RMSProp to tackle a certain counterexample and train deep neural networks. Numerical results are exactly in accord with our theoretical analysis.

Deep Convolutional Neural Networks (CNNs) have been pushing the frontier of face recognition over past years. However, existing CNN models are far less accurate when handling partially occluded faces. These general face models generalize poorly for occlusions on variable facial areas. Inspired by the fact that human visual system explicitly ignores the occlusion and only focuses on the non-occluded facial areas, we propose a mask learning strategy to find and discard corrupted feature elements from recognition. A mask dictionary is firstly established by exploiting the differences between the top conv features of occluded and occlusion-free face pairs using innovatively designed pairwise differential siamese network (PDSN). Each item of this dictionary captures the correspondence between occluded facial areas and corrupted feature elements, which is named Feature Discarding Mask (FDM). When dealing with a face image with random partial occlusions, we generate its FDM by combining relevant dictionary items and then multiply it with the original features to eliminate those corrupted feature elements from recognition. Comprehensive experiments on both synthesized and realistic occluded face datasets show that the proposed algorithm significantly outperforms the state-of-the-art systems.

In this paper, we propose to guide the video caption generation with Part-of-Speech (POS) information, based on a gated fusion of multiple representations of input videos. We construct a novel gated fusion network, with one particularly designed cross-gating (CG) block, to effectively encode and fuse different types of representations, e.g., the motion and content features of an input video. One POS sequence generator relies on this fused representation to predict the global syntactic structure, which is thereafter leveraged to guide the video captioning generation and control the syntax of the generated sentence. Specifically, a gating strategy is proposed to dynamically and adaptively incorporate the global syntactic POS information into the decoder for generating each word. Experimental results on two benchmark datasets, namely MSR-VTT and MSVD, demonstrate that the proposed model can well exploit complementary information from multiple representations, resulting in improved performances. Moreover, the generated global POS information can well capture the global syntactic structure of the sentence, and thus be exploited to control the syntactic structure of the description. Such POS information not only boosts the video captioning performance but also improves the diversity of the generated captions. Our code is at: https://github.com/vsislab/Controllable_XGating.

With the rapid developments of deep neural networks, numerous deep cross-modal analysis methods have been presented and are being applied in widespread real-world applications, including healthcare and safety-critical environments. However, the recent studies on robustness and stability of deep neural networks show that a microscopic modification, known as adversarial sample, which is even imperceptible to humans, can easily fool a well-performed deep neural network and brings a new obstacle to deep cross-modal correlation exploring. In this paper, we propose a novel Cross-Modal correlation Learning with Adversarial samples, namely CMLA, which for the first time presents the existence of adversarial samples in cross-modal data. Moreover, we provide a simple yet effective adversarial sample learning method, where inter- and intra- modality similarity regularizations across different modalities are simultaneously integrated into the learning of adversarial samples. Finally, our proposed CMLA is demonstrated to be highly effective in cross-modal hashing based retrieval. Extensive experiments on two cross-modal benchmark datasets show that the adversarial examples produced by our CMLA are efficient in fooling a target deep cross-modal hashing network. On the other hand, such adversarial examples can significantly strengthen the robustness of the target network by conducting an adversarial training.

Temporal sentence grounding in videos aims to detect and localize one target video segment, which semantically corresponds to a given sentence. Existing methods mainly tackle this task via matching and aligning semantics between a sentence and candidate video segments, while neglect the fact that the sentence information plays an important role in temporally correlating and composing the described contents in videos. In this paper, we propose a novel semantic conditioned dynamic modulation (SCDM) mechanism, which relies on the sentence semantics to modulate the temporal convolution operations for better correlating and composing the sentence related video contents over time. More importantly, the proposed SCDM performs dynamically with respect to the diverse video contents so as to establish a more precise matching relationship between sentence and video, thereby improving the temporal grounding accuracy. Extensive experiments on three public datasets demonstrate that our proposed model outperforms the state-of-the-arts with clear margins, illustrating the ability of SCDM to better associate and localize relevant video contents for temporal sentence grounding. Our code for this paper is available at https://github.com/yytzsy/SCDM.

Unsupervised domain adaptation (UDA) aims to enhance the generalization capability of a certain model from a source domain to a target domain. UDA is of particular significance since no extra effort is devoted to annotating target domain samples. However, the different data distributions in the two domains, or \emph{domain shift/discrepancy}, inevitably compromise the UDA performance. Although there has been a progress in matching the marginal distributions between two domains, the classifier favors the source domain features and makes incorrect predictions on the target domain due to category-agnostic feature alignment. In this paper, we propose a novel category anchor-guided (CAG) UDA model for semantic segmentation, which explicitly enforces category-aware feature alignment to learn shared discriminative features and classifiers simultaneously. First, the category-wise centroids of the source domain features are used as guided anchors to identify the active features in the target domain and also assign them pseudo-labels. Then, we leverage an anchor-based pixel-level distance loss and a discriminative loss to drive the intra-category features closer and the inter-category features further apart, respectively. Finally, we devise a stagewise training mechanism to reduce the error accumulation and adapt the proposed model progressively. Experiments on both the GTA5$\rightarrow $Cityscapes and SYNTHIA$\rightarrow $Cityscapes scenarios demonstrate the superiority of our CAG-UDA model over the state-of-the-art methods. The code is available at \url{https://github.com/RogerZhangzz/CAG\_UDA}.

Sparse support vector machine (SVM) is a popular classification technique that can simultaneously learn a small set of the most interpretable features and identify the support vectors. It has achieved great successes in many real-world applications. However, for large-scale problems involving a huge number of samples and ultra-high dimensional features, solving sparse SVMs remains challenging. By noting that sparse SVMs induce sparsities in both feature and sample spaces, we propose a novel approach, which is based on accurate estimations of the primal and dual optima of sparse SVMs, to simultaneously identify the inactive features and samples that are guaranteed to be irrelevant to the outputs. Thus, we can remove the identified inactive samples and features from the training phase, leading to substantial savings in the computational cost without sacrificing the accuracy. Moreover, we show that our method can be extended to multi-class sparse support vector machines. To the best of our knowledge, the proposed method is the first static feature and sample reduction method for sparse SVMs and multi-class sparse SVMs. Experiments on both synthetic and real data sets demonstrate that our approach significantly outperforms state-of-the-art methods and the speedup gained by our approach can be orders of magnitude.

Hashing plays a pivotal role in nearest-neighbor searching for large-scale image retrieval. Recently, deep learning-based hashing methods have achieved promising performance. However, most of these deep methods involve discriminative models, which require large-scale, labeled training datasets, thus hindering their real-world applications. In this paper, we propose a novel strategy to exploit the semantic similarity of the training data and design an efficient generative adversarial framework to learn binary hash codes in an unsupervised manner. Specifically, our model consists of three different neural networks: an encoder network to learn hash codes from images, a generative network to generate images from hash codes, and a discriminative network to distinguish between pairs of hash codes and images. By adversarially training these networks, we successfully learn mutually coherent encoder and generative networks, and can output efficient hash codes from the encoder network. We also propose a novel strategy, which utilizes both feature and neighbor similarities, to construct a semantic similarity matrix, then use this matrix to guide the hash code learning process. Integrating the supervision of this semantic similarity matrix into the adversarial learning framework can efficiently preserve the semantic information of training data in Hamming space. The experimental results on three widely used benchmarks show that our method not only significantly outperforms several state-of-the-art unsupervised hashing methods, but also achieves comparable performance with popular supervised hashing methods.

In this paper, we propose the novel principal backpropagation networks (PBNets) to revisit the backpropagation algorithms commonly used in training two-stream networks for video action recognition. We content that existing approaches always take all the frames/snippets for the backpropagation not optimal for video recognition since the desired actions only occur in a short period within a video. To remedy these drawbacks, we design a watch-and-choose mechanism. In particular, the watching stage exploits a dense snippet-wise temporal pooling strategy to discover the global characteristic for each input video, while the choosing phase only backpropagates a small number of representative snippets that are selected with two novel strategies, i.e., Max-rule and KL-rule. We prove that with the proposed selection strategies, performing the backpropagation on the selected subset is capable of decreasing the loss of the whole snippets as well. The proposed PBNets are evaluated on two standard video action recognition benchmarks UCF101 and HMDB51, where it surpasses the state of the arts consistently, but requiring less memory and computation to achieve high performance.

Recent research have shown the potential of using convolutional neural networks (CNNs) to accomplish single image dehazing. In this work, we take one step further to explore the possibility of exploiting a network to perform haze removal for videos. Unlike single image dehazing, video based approaches can take advantage of the abundant information that exists across neighboring frames. In this work, assuming that a scene point yields highly correlated transmission values between adjacent video frames, we develop a deep learning solution for video dehazing, where a CNN is trained end-to-end to learn how to accumulate information across frames for transmission estimation. The estimated transmission map is subsequently used to recover a haze-free frame via atmospheric scattering model. In addition, as the semantic information of a scene provides a strong prior for image restoration, we propose to incorporate global semantic priors as input to regularize the transmission maps so that the estimated maps can be smooth in the regions of the same object and only discontinuous across the boundaries of different objects. To train this network, we generate a dataset consisted of synthetic hazy and haze-free videos for supervision based on the NYU depth dataset. We show that the features learned from this dataset are capable of removing haze that arises in outdoor scenes in a wide range of videos. Extensive experiments demonstrate that the proposed algorithm performs favorably against the state-of-the-art methods on both synthetic and real-world videos.

Camera shake or target movement often leads to undesired blur effects in videos captured by a hand-held camera. Despite significant efforts having been devoted to video-deblur research, two major challenges remain: 1) how to model the spatio-temporal characteristics across both the spatial domain (i.e., image plane) and the temporal domain (i.e., neighboring frames) and 2) how to restore sharp image details with respect to the conventionally adopted metric of pixel-wise errors. In this paper, to address the first challenge, we propose a deblurring network (DBLRNet) for spatial-temporal learning by applying a 3D convolution to both the spatial and temporal domains. Our DBLRNet is able to capture jointly spatial and temporal information encoded in neighboring frames, which directly contributes to the improved video deblur performance. To tackle the second challenge, we leverage the developed DBLRNet as a generator in the generative adversarial network (GAN) architecture and employ a content loss in addition to an adversarial loss for efficient adversarial training. The developed network, which we name as deblurring GAN, is tested on two standard benchmarks and achieves the state-of-the-art performance.

Domain adaptation is the situation for supervised learning in which the training data are sampled from the source domain while the test data are sampled from the target domain that follows a different distribution. The key to solving such a problem is to reduce effects of the discrepancy between the training data and test data. Recently, deep learning methods that employ stacked denoising auto-encoders (SDAs) to learn new representations for both domains have been successfully applied in domain adaptation. And, remarkable performance on multi-domain sentiment analysis datasets has been reported, making deep learning a promising approach to domain adaptation problems. In this paper, a deep learning method called Stacked Robust Adaptively Regularized Auto-regressions (SRARAs) is proposed to learn useful representations for domain adaptation problems. Each layer of SRARAs contains two steps: a linear transformation step, which is based on robust adaptively regularized auto-regression, and a non-linear squashing transformation step. The first step aims at reducing the discrepancy between the training data and test data, and the second step is to introduce non-linearity and control the range of the elements in the outputs. The experimental results on text and image datasets demonstrate that the proposed method is very effective.

With the availability of abundant online multi-relational video information, recommender systems that can effectively exploit these sorts of data and suggest creatively interesting items will become increasingly important. Recent research illustrates that tensor models offer effective approaches for complex multi-relational data learning and missing element completion. So far, most tensor-based user clustering models have focused on the accuracy of recommendation. Given the dynamic nature of online media, recommendation in this setting is more challenging as it is difficult to capture the users’ dynamic topic distributions in sparse data settings as well as to identify unseen items as candidates of recommendation. Targeting at constructing a recommender system that can encourage more creativity, a deep Bayesian probabilistic tensor framework for tag and item recommendation is proposed. During the score ranking processes, a metric called Bayesian surprise is incorporated to increase the creativity of the recommended candidates. The new algorithm, called Deep Canonical PARAFAC Factorization (DCPF), is evaluated on both synthetic and large-scale real-world problems. An empirical study for video recommendation demonstrates the superiority of the proposed model and indicates that it can better capture the latent patterns of interactions and generates interesting recommendations based on creative tag combinations.

Recently, much advance has been made in image captioning, and an encoder-decoder framework has achieved outstanding performance for this task. In this paper, we propose an extension of the encoder-decoder framework by adding a component called guiding network. The guiding network models the attribute properties of input images, and its output is leveraged to compose the input of the decoder at each time step. The guiding network can be plugged into the current encoder-decoder framework and trained in an end-to-end manner. Hence, the guiding vector can be adaptively learned according to the signal from the decoder, making itself to embed information from both image and language. Additionally, discriminative supervision can be employed to further improve the quality of guidance. The advantages of our proposed approach are verified by experiments carried out on the MS COCO dataset.

Learning representation from relative similarity comparisons, often called ordinal embedding, gains rising attention in recent years. Most of the existing methods are batch methods designed mainly based on the convex optimization, say, the projected gradient descent method. However, they are generally time-consuming due to that the singular value decomposition (SVD) is commonly adopted during the update, especially when the data size is very large. To overcome this challenge, we propose a stochastic algorithm called SVRG-SBB, which has the following features: (a) SVD-free via dropping convexity, with good scalability by the use of stochastic algorithm, i.e., stochastic variance reduced gradient (SVRG), and (b) adaptive step size choice via introducing a new stabilized Barzilai-Borwein (SBB) method as the original version for convex problems might fail for the considered stochastic non-convex optimization problem. Moreover, we show that the proposed algorithm converges to a stationary point at a rate O(1/T) in our setting, where T is the number of total iterations. Numerous simulations and real-world data experiments are conducted to show the effectiveness of the proposed algorithm via comparing with the state-of-the-art methods, particularly, much lower computational cost with good prediction performance.

This paper addresses the problem of video object segmentation, where the initial object mask is given in the first frame of an input video. We propose a novel spatiotemporal Markov Random Field (MRF) model defined over pixels to handle this problem. Unlike conventional MRF models, the spatial dependencies among pixels in our model are encoded by a Convolutional Neural Network (CNN). Specifically, for a given object, the probability of a labeling to a set of spatially neighboring pixels can be predicted by a CNN trained for this specific object. As a result, higher-order, richer dependencies among pixels in the set can be implicitly modeled by the CNN. With temporal dependencies established by optical flow, the resulting MRF model combines both spatial and temporal cues for tackling video object segmentation. However, performing inference in the MRF model is very difficult due to the very high-order dependencies. To this end, we propose a novel CNN-embedded algorithm to perform approximate inference in the MRF. This algorithm proceeds by alternating between a temporal fusion step and a feed-forward CNN step. When initialized with an appearance-based one-shot segmentation CNN, our model outperforms the winning entries of the DAVIS 2017 Challenge, without resorting to model ensembling or any dedicated detectors.

Recently, caption generation with an encoder-decoder framework has been extensively studied and applied in different domains, such as image captioning, code captioning, and so on. In this paper, we propose a novel architecture, namely Auto-Reconstructor Network (ARNet), which, coupling with the conventional encoder-decoder framework, works in an end-to-end fashion to generate captions. ARNet aims at reconstructing the previous hidden state with the present one, besides behaving as the input-dependent transition operator. Therefore, ARNet encourages the current hidden state to embed more information from the previous one, which can help regularize the transition dynamics of recurrent neural networks (RNNs). Extensive experimental results show that our proposed ARNet boosts the performance over the existing encoder-decoder models on both image captioning and source code captioning tasks. Additionally, ARNet remarkably reduces the discrepancy between training and inference processes for caption generation. Furthermore, the performance on permuted sequential MNIST demonstrates that ARNet can effectively regularize RNN, especially on modeling long-term dependencies. Our code is available at: https://github.com/chenxinpeng/ARNet.

We propose a novel framework called Semantics-Preserving Adversarial Embedding Network (SP-AEN) for zero-shot visual recognition (ZSL), where test images and their classes are both unseen during training. SP-AEN aims to tackle the inherent problem - semantic loss - in the prevailing family of embedding-based ZSL, where some semantics would be discarded during training if they are non-discriminative for training classes, but could become critical for recognizing test classes. Specifically, SP-AEN prevents the semantic loss by introducing an independent visual-to-semantic space embedder which disentangles the semantic space into two subspaces for the two arguably conflicting objectives: classification and reconstruction. Through adversarial learning of the two subspaces, SP-AEN can transfer the semantics from the reconstructive subspace to the discriminative one, accomplishing the improved zero-shot recognition of unseen classes. Comparing with prior works, SP-AEN can not only improve classification but also generate photo-realistic images, demonstrating the effectiveness of semantic preservation. On four popular benchmarks: CUB, AWA, SUN and aPY, SP-AEN considerably outperforms other state-of-the-art methods by an absolute performance difference of 12.2%, 9.3%, 4.0% and 3.6% in terms of harmonic mean values [62].

Inspired by the recent neuroscience studies on the left-right asymmetry of the human brain in processing low and high spatial frequency information, this paper introduces a dual skipping network which carries out coarse-to-fine object categorization. Such a network has two branches to simultaneously deal with both coarse and fine-grained classification tasks. Specifically, we propose a layer-skipping mechanism that learns a gating network to predict which layers to skip in the testing stage. This layer-skipping mechanism endows the network with good flexibility and capability in practice. Evaluations are conducted on several widely used coarse-to-fine object categorization benchmarks, and promising results are achieved by our proposed network model.

Leveraging the disparity information from both left and right views is crucial for stereo disparity estimation. Left-right consistency check is an effective way to enhance the disparity estimation by referring to the information from the opposite view. However, the conventional left-right consistency check is an isolated post-processing step and heavily hand-crafted. This paper proposes a novel left-right comparative recurrent model to perform left-right consistency checking jointly with disparity estimation. At each recurrent step, the model produces disparity results for both views, and then performs online left-right comparison to identify the mismatched regions which may probably contain erroneously labeled pixels. A soft attention mechanism is introduced, which employs the learned error maps for better guiding the model to selectively focus on refining the unreliable regions at the next recurrent step. In this way, the generated disparity maps are progressively improved by the proposed recurrent model. Extensive evaluations on KITTI 2015, Scene Flow and Middlebury benchmarks validate the effectiveness of our model, demonstrating that state-of-the-art stereo disparity estimation results can be achieved by this new model.

Thanks to the success of deep learning, cross-modal retrieval has made significant progress recently. However, there still remains a crucial bottleneck: how to bridge the modality gap to further enhance the retrieval accuracy. In this paper, we propose a self-supervised adversarial hashing (SSAH) approach, which lies among the early attempts to incorporate adversarial learning into cross-modal hashing in a self-supervised fashion. The primary contribution of this work is that two adversarial networks are leveraged to maximize the semantic correlation and consistency of the representations between different modalities. In addition, we harness a self-supervised semantic network to discover high-level semantic information in the form of multi-label annotations. Such information guides the feature learning process and preserves the modality relationships in both the common semantic space and the Hamming space. Extensive experiments carried out on three benchmark datasets validate that the proposed SSAH surpasses the state-of-the-art methods.

In this paper, we propose an efficient algorithm to directly restore a clear image from a hazy input. The proposed algorithm hinges on an end-to-end trainable neural network that consists of an encoder and a decoder. The encoder is exploited to capture the context of the derived input images, while the decoder is employed to estimate the contribution of each input to the final dehazed result using the learned representations attributed to the encoder. The constructed network adopts a novel fusion-based strategy which derives three inputs from an original hazy image by applying White Balance (WB), Contrast Enhancing (CE), and Gamma Correction (GC). We compute pixel-wise confidence maps based on the appearance differences between these different inputs to blend the information of the derived inputs and preserve the regions with pleasant visibility. The final dehazed image is yielded by gating the important features of the derived inputs. To train the network, we introduce a multi-scale approach such that the halo artifacts can be avoided. Extensive experimental results on both synthetic and real-world images demonstrate that the proposed algorithm performs favorably against the state-of-the-art algorithms.

In this paper, the problem of describing visual contents of a video sequence with natural language is addressed. Unlike previous video captioning work mainly exploiting the cues of video contents to make a language description, we propose a reconstruction network (RecNet) with a novel encoder-decoder-reconstructor architecture, which leverages both the forward (video to sentence) and backward (sentence to video) flows for video captioning. Specifically, the encoder-decoder makes use of the forward flow to produce the sentence description based on the encoded video semantic features. Two types of reconstructors are customized to employ the backward flow and reproduce the video features based on the hidden state sequence generated by the decoder. The generation loss yielded by the encoder-decoder and the reconstruction loss introduced by the reconstructor are jointly drawn into training the proposed RecNet in an end-to-end fashion. Experimental results on benchmark datasets demonstrate that the proposed reconstructor can boost the encoder-decoder models and leads to significant gains in video caption accuracy.

Dense video captioning is a newly emerging task that aims at both localizing and describing all events in a video. We identify and tackle two challenges on this task, namely, (1) how to utilize both past and future contexts for accurate event proposal predictions, and (2) how to construct informative input to the decoder for generating natural event descriptions. First, previous works predominantly generate temporal event proposals in the forward direction, which neglects future video context. We propose a bidirectional proposal method that effectively exploits both past and future contexts to make proposal predictions. Second, different events ending at (nearly) the same time are indistinguishable in the previous works, resulting in the same captions. We solve this problem by representing each event with an attentive fusion of hidden states from the proposal module and video contents (e.g., C3D features). We further propose a novel context gating mechanism to balance the contributions from the current event and its surrounding contexts dynamically. We empirically show that our attentively fused event representation is superior to the proposal hidden states or video contents alone. By coupling proposal and captioning modules into one unified framework, our model outperforms the state-of-the-arts on the ActivityNet Captions dataset with a relative gain of over 100% (Meteor score increases from 4.82 to 9.65).

Face recognition has made extraordinary progress owing to the advancement of deep convolutional neural networks (CNNs). The central task of face recognition, including face verification and identification, involves face feature discrimination. However, the traditional softmax loss of deep CNNs usually lacks the power of discrimination. To address this problem, recently several loss functions such as center loss, large margin softmax loss, and angular softmax loss have been proposed. All these improved losses share the same idea: maximizing inter-class variance and minimizing intra-class variance. In this paper, we propose a novel loss function, namely large margin cosine loss (LMCL), to realize this idea from a different perspective. More specifically, we reformulate the softmax loss as a cosine loss by L2 normalizing both features and weight vectors to remove radial variations, based on which a cosine margin term is introduced to further maximize the decision margin in the angular space. As a result, minimum intra-class variance and maximum inter-class variance are achieved by virtue of normalization and cosine decision margin maximization. We refer to our model trained with LMCL as CosFace. Extensive experimental evaluations are conducted on the most popular public-domain face recognition datasets such as MegaFace Challenge, Youtube Faces (YTF) and Labeled Face in the Wild (LFW). We achieve the state-of-the-art performance on these benchmarks, which confirms the effectiveness of our proposed approach.

In this work we propose a new automatic image annotation model, dubbed diverse and distinct image annotation (D2IA). The generative model D2IA is inspired by the ensemble of human annotations, which create semantically relevant, yet distinct and diverse tags. In D2IA, we generate a relevant and distinct tag subset, in which the tags are relevant to the image contents and semantically distinct to each other, using sequential sampling from a determinantal point process (DPP) model. Multiple such tag subsets that cover diverse semantic aspects or diverse semantic levels of the image contents are generated by randomly perturbing the DPP sampling process. We leverage a generative adversarial network (GAN) model to train D2IA. Extensive experiments including quantitative and qualitative comparisons, as well as human subject studies, on two benchmark datasets demonstrate that the proposed model can produce more diverse and distinct tags than the state-of-the-arts.

Taking a photo outside, can we predict the immediate future, e.g., how would the cloud move in the sky? We address this problem by presenting a generative adversarial network (GAN) based two-stage approach to generating realistic time-lapse videos of high resolution. Given the first frame, our model learns to generate long-term future frames. The first stage generates videos of realistic contents for each frame. The second stage refines the generated video from the first stage by enforcing it to be closer to real videos with regard to motion dynamics. To further encourage vivid motion in the final generated video, Gram matrix is employed to model the motion more precisely. We build a large scale time-lapse dataset, and test our approach on this new dataset. Using our model, we are able to generate realistic videos of up to 128 Ã- 128 resolution for 32 frames. Quantitative and qualitative experiment results demonstrate the superiority of our model over the state-of-the-art models.

We study active object tracking, where a tracker takes as input the visual observation (i.e. frame sequence) and produces the camera control signal (e.g., move forward, turn left, etc). Conventional methods tackle the tracking and the camera control separately, which is challenging to tune jointly. It also incurs many human efforts for labeling and many expensive trial-and-errors in real-world. To address these issues, we propose, in this paper, an end-to-end solution via deep reinforcement learning, where a ConvNet-LSTM function approximator is adopted for the direct frame-to-action prediction. We further propose an environment augmentation technique and a customized reward function, which are crucial for a successful training. The tracker trained in simulators (ViZDoom, Unreal Engine) shows good generalization in the case of unseen object moving path, unseen object appearance, unseen background, and distracting object. It can restore tracking when occasionally losing the target. With the experiments over the VOT dataset, we also find that the tracking ability, obtained solely from simulators, can potentially transfer to real-world scenarios.

We propose a novel algorithmic framework of Variable Metric Over-Relaxed Hybrid Proximal Extra-gradient (VMOR-HPE) method with a global convergence guarantee for the maximal monotone operator inclusion problem. Its iteration complexities and local linear convergence rate are provided, which theoretically demonstrate that a large over-relaxed step-size contributes to accelerating the proposed VMOR-HPE as a byproduct. Specifically, we find that a large class of primal and primal-dual operator splitting algorithms are all special cases of VMOR-HPE. Hence, the proposed framework offers a new insight into these operator splitting algorithms. In addition, we apply VMOR-HPE to the Karush-Kuhn-Tucker (KKT) generalized equation of linear equality constrained multi-block composite convex optimization, yielding a new algorithm, namely nonsymmetric Proximal Alternating Direction Method of Multipliers with a preconditioned Extra-gradient step in which the preconditioned metric is generated by a blockwise Barzilai-Borwein line search technique (PADMM-EBB). We also establish iteration complexities of PADMM-EBB in terms of the KKT residual. Finally, we apply PADMM-EBB to handle the nonnegative dual graph regularized low-rank representation problem. Promising results on synthetic and real datasets corroborate the efficacy of PADMM-EBB.

Submodular functions are discrete analogs of convex functions, which have applications in various fields, including machine learning and computer vision. However, in large-scale applications, solving Submodular Function Minimization (SFM) problems remains challenging. In this paper, we make the first attempt to extend the emerging technique named screening in large-scale sparse learning to SFM for accelerating its optimization process. We first conduct a careful studying of the relationships between SFM and the corresponding convex proximal problems, as well as the accurate primal optimum estimation of the proximal problems. Relying on this study, we subsequently propose a novel safe screening method to quickly identify the elements guaranteed to be included (we refer to them as active) or excluded (inactive) in the final optimal solution of SFM during the optimization process. By removing the inactive elements and fixing the active ones, the problem size can be dramatically reduced, leading to great savings in the computational cost without sacrificing any accuracy. To the best of our knowledge, the proposed method is the first screening method in the fields of SFM and even combinatorial optimization, thus pointing out a new direction for accelerating SFM algorithms. Experiment results on both synthetic and real datasets demonstrate the significant speedups gained by our approach.

Human motion prediction aims at generating future frames of human motion based on an observed sequence of skeletons. Recent methods employ the latest hidden states of a recurrent neural network (RNN) to encode the historical skeletons, which can only address short-term prediction. In this work, we propose a motion context modeling by summarizing the historical human motion with respect to the current prediction. A modified highway unit (MHU) is proposed for efficiently eliminating motionless joints and estimating next pose given the motion context. Furthermore, we enhance the motion dynamic by minimizing the gram matrix loss for long-term motion prediction. Experimental results show that the proposed model can promisingly forecast the human future movements, which yields superior performances over related state-of-the-art approaches. Moreover, specifying the motion context with the activity labels enables our model to perform human motion transfer.

In this paper, we propose a deep learning approach to tackle the automatic summarization tasks by incorporating topic information into the convolutional sequence-to-sequence (ConvS2S) model and using self-critical sequence training (SCST) for optimization. Through jointly attending to topics and word-level alignment, our approach can improve coherence, diversity, and informativeness of generated summaries via a biased probability generation mechanism. On the other hand, reinforcement training, like SCST, directly optimizes the proposed model with respect to the non-differentiable metric ROUGE, which also avoids the exposure bias during inference. We carry out the experimental evaluation with state-of-the-art methods over the Gigaword, DUC-2004, and LCSTS datasets. The empirical results demonstrate the superiority of our proposed method in the abstractive summarization.

Hashing is becoming increasingly popular for approximate nearest neighbor searching in massive databases due to its storage and search efficiency. Recent supervised hashing methods, which usually construct semantic similarity matrices to guide hash code learning using label information, have shown promising results. However, it is relatively difficult to capture and utilize the semantic relationships between points in unsupervised settings. To address this problem, we propose a novel unsupervised deep framework called Semantic Structure-based unsupervised Deep Hashing (SSDH). We first empirically study the deep feature statistics, and find that the distribution of the cosine distance for point pairs can be estimated by two half Gaussian distributions. Based on this observation, we construct the semantic structure by considering points with distances obviously smaller than the others as semantically similar and points with distances obviously larger than the others as semantically dissimilar. We then design a deep architecture and a pair-wise loss function to preserve this semantic structure in Hamming space. Extensive experiments show that SSDH significantly outperforms current state-of-the-art methods.

In this paper, we present a deep convolutional neural network to capture the inherent properties of image degradation, which can handle different kernels and saturated pixels in a unified framework. The proposed neural network is motivated by the low-rank property of pseudo-inverse kernels. We first compute a generalized low-rank approximation for a large number of blur kernels, and then use separable filters to initialize the convolutional parameters in the network. Our analysis shows that the estimated decomposed matrices contain the most essential information of the input kernel, which ensures the proposed network to handle various blurs in a unified framework and generate high-quality deblurring results. Experimental results on benchmark datasets with noise and saturated pixels demonstrate that the proposed algorithm performs favorably against state-of-the-art methods.

Nonlocal neural networks have been proposed and shown to be effective in several computer vision tasks, where the nonlocal operations can directly capture long-range dependencies in the feature space. In this paper, we study the nature of diffusion and damping effect of nonlocal networks by doing spectrum analysis on the weight matrices of the well-trained networks, and then propose a new formulation of the nonlocal block. The new block not only learns the nonlocal interactions but also has stable dynamics, thus allowing deeper nonlocal structures. Moreover, we interpret our formulation from the general nonlocal modeling perspective, where we make connections between the proposed nonlocal network and other nonlocal models, such as nonlocal diffusion process and Markov jump process.

We propose a novel Wasserstein method with a distillation mechanism, yielding joint learning of word embeddings and topics. The proposed method is based on the fact that the Euclidean distance between word embeddings may be employed as the underlying distance in the Wasserstein topic model. The word distributions of topics, their optimal transport to the word distributions of documents, and the embeddings of words are learned in a unified framework. When learning the topic model, we leverage a distilled ground-distance matrix to update the topic distributions and smoothly calculate the corresponding optimal transports. Such a strategy provides the updating of word embeddings with robust guidance, improving algorithm convergence. As an application, we focus on patient admission records, in which the proposed method embeds the codes of diseases and procedures and learns the topics of admissions, obtaining superior performance on clinically-meaningful disease network construction, mortality prediction as a function of admission codes, and procedure recommendation.

We propose a parsimonious quantile regression framework to learn the dynamic tail behaviors of financial asset returns. Our model captures well both the time-varying characteristic and the asymmetrical heavy-tail property of financial time series. It combines the merits of a popular sequential neural network model, i.e., LSTM, with a novel parametric quantile function that we construct to represent the conditional distribution of asset returns. Our model also captures individually the serial dependences of higher moments, rather than just the volatility. Across a wide range of asset classes, the out-of-sample forecasts of conditional quantiles or VaR of our model outperform the GARCH family. Further, the proposed approach does not suffer from the issue of quantile crossing, nor does it expose to the ill-posedness comparing to the parametric probability density function approach.

Graph matching has received persistent attention over decades, which can be formulated as a quadratic assignment problem (QAP). We show that a large family of functions, which we define as Separable Functions, can approximate discrete graph matching in the continuous domain asymptotically by varying the approximation controlling parameters. We also study the properties of global optimality and devise convex/concave-preserving extensions to the widely used Lawler's QAP form. Our theoretical findings show the potential for deriving new algorithms and techniques for graph matching. We deliver solvers based on two specific instances of Separable Functions, and the state-of-the-art performance of our method is verified on popular benchmarks.

Recently, the booming fashion sector and its huge potential benefits have attracted tremendous attention from many research communities. In particular, increasing research efforts have been dedicated to the complementary clothing matching as matching clothes to make a suitable outfit has become a daily headache for many people, especially those who do not have the sense of aesthetics. Thanks to the remarkable success of neural networks in various applications such as the image classification and speech recognition, the researchers are enabled to adopt the data-driven learning methods to analyze fashion items. Nevertheless, existing studies overlook the rich valuable knowledge (rules) accumulated in fashion domain, especially the rules regarding clothing matching. Towards this end, in this work, we shed light on the complementary clothing matching by integrating the advanced deep neural networks and the rich fashion domain knowledge. Considering that the rules can be fuzzy and different rules may have different confidence levels to different samples, we present a neural compatibility modeling scheme with attentive knowledge distillation based on the teacher-student network scheme. Extensive experiments on the real-world dataset show the superiority of our model over several state-of-the-art methods. Based upon the comparisons, we observe certain fashion insights that can add value to the fashion matching study. As a byproduct, we released the codes, and involved parameters to benefit other researchers.

The booming of social networks has given rise to a large volume of user-generated contents (UGCs), most of which are free and publicly available. A lot of users' personal aspects can be extracted from these UGCs to facilitate personalized applications as validated by many previous studies. Despite their value, UGCs can place users at high privacy risks, which thus far remains largely untapped. Privacy is defined as the individual's ability to control what information is disclosed, to whom, when and under what circumstances. As people and information both play significant roles, privacy has been elaborated as a boundary regulation process, where individuals regulate interaction with others by altering the openness degree of themselves to others. In this paper, we aim to reduce users' privacy risks on social networks by answering the question of Who Can See What. Towards this goal, we present a novel scheme, comprising of descriptive, predictive and prescriptive components. In particular, we first collect a set of posts and extract a group of privacy-oriented features to describe the posts. We then propose a novel taxonomy-guided multi-task learning model to predict which personal aspects are uncovered by the posts. Lastly, we construct standard guidelines by the user study with 400 users to regularize users' actions for preventing their privacy leakage. Extensive experiments on a real-world dataset well verified our scheme.

Recently with the explosive growth of visual content on the Internet, large-scale image search has attracted intensive attention. It has been shown that mapping high-dimensional image descriptors to compact binary codes can lead to considerable efficiency gains in both storage and performing similarity computation of images. However, most existing methods still suffer from expensive training devoted to large-scale binary code learning. To address this issue, we propose a sub-selection based matrix manipulation algorithm, which can significantly reduce the computational cost of code learning. As case studies, we apply the sub-selection algorithm to several popular quantization techniques including cases using linear and nonlinear mappings. Crucially, we can justify the resulting sub-selective quantization by proving its theoretic properties. Extensive experiments are carried out on three image benchmarks with up to one million samples, corroborating the efficacy of the sub-selective quantization method in terms of image retrieval.

Hashing is emerging as a powerful tool for building highly efficient indices in large-scale search systems. In this paper, we study spectral hashing (SH), which is a classical method of unsupervised hashing. In general, SH solves for the hash codes by minimizing an objective function that tries to preserve the similarity structure of the data given. Although computationally simple, very often SH performs unsatisfactorily and lags distinctly behind the state-of-the-art methods. We observe that the inferior performance of SH is mainly due to its imperfect formulation; that is, the optimization of the minimization problem in SH actually cannot ensure that the similarity structure of the high-dimensional data is really preserved in the low-dimensional hash code space. In this paper, we, therefore, introduce reversed SH (ReSH), which is SH with its input and output interchanged. Unlike SH, which estimates the similarity structure from the given high-dimensional data, our ReSH defines the similarities between data points according to the unknown low-dimensional hash codes. Equipped with such a reversal mechanism, ReSH can seamlessly overcome the drawback of SH. More precisely, the minimization problem in our ReSH can be optimized if and only if similar data points are mapped to adjacent hash codes, and mostly important, dissimilar data points are considerably separated from each other in the code space. Finally, we solve the minimization problem in ReSH by multilayer neural networks and obtain state-of-the-art retrieval results on three benchmark data sets.

With explosive growth of data volume and ever-increasing diversity of data modalities, cross-modal similarity search, which conducts nearest neighbor search across different modalities, has been attracting increasing interest. This paper presents a deep compact code learning solution for efficient cross-modal similarity search. Many recent studies have proven that quantization-based approaches perform generally better than hashing-based approaches on single-modal similarity search. In this paper, we propose a deep quantization approach, which is among the early attempts of leveraging deep neural networks into quantization-based cross-modal similarity search. Our approach, dubbed shared predictive deep quantization (SPDQ), explicitly formulates a shared subspace across different modalities and two private subspaces for individual modalities, and representations in the shared subspace and the private subspaces are learned simultaneously by embedding them to a reproducing kernel Hilbert space, where the mean embedding of different modality distributions can be explicitly compared. In addition, in the shared subspace, a quantizer is learned to produce the semantics preserving compact codes with the help of label alignment. Thanks to this novel network architecture in cooperation with supervised quantization training, SPDQ can preserve intramodal and intermodal similarities as much as possible and greatly reduce quantization error. Experiments on two popular benchmarks corroborate that our approach outperforms state-of-the-art methods.

Features extracted by deep networks have been popular in many visual search tasks. This article studies deep network structures and training schemes for mobile visual search. The goal is to learn an effective yet portable feature representation that is suitable for bridging the domain gap between mobile user photos and (mostly) professionally taken product images while keeping the computational cost acceptable for mobile-based applications. The technical contributions are twofold. First, we propose an alternative of the contrastive loss popularly used for training deep Siamese networks, namely robust contrastive loss, where we relax the penalty on some positive and negative pairs to alleviate overfitting. Second, a simple multitask fine-tuning scheme is leveraged to train the network, which not only utilizes knowledge from the provided training photo pairs but also harnesses additional information from the large ImageNet dataset to regularize the fine-tuning process. Extensive experiments on challenging real-world datasets demonstrate that both the robust contrastive loss and the multitask fine-tuning scheme are effective, leading to very promising results with a time cost suitable for mobile product search scenarios.

We propose a fast first-order method to solve multi-term nonsmooth composite convex minimization problems by employing a recent proximal average approximation technique and a novel adaptive parameter tuning technique. Thanks to this powerful parameter tuning technique, the proximal gradient step can be performed with a much larger stepsize in the algorithm implementation compared with the prior PA-APG method, which is the core to enable significant improvements in practical performance. Moreover, by choosing the approximation parameter adaptively, the proposed method is shown to enjoy the O(1/k) iteration complexity theoretically without needing any extra computational cost, while the PA-APG method incurs much more iterations for convergence. The preliminary experimental results on overlapping group Lasso and graph-guided fused Lasso problems confirm our theoretic claim well, and indicate that the proposed method is almost five times faster than the state-of-the-art PA-APG method and therefore suitable for higher-precision required optimization.

With benefits of low storage cost and fast query speed, cross-modal hashing has received considerable attention recently. However, almost all existing methods on cross-modal hashing cannot obtain powerful hash codes due to directly utilizing hand-crafted features or ignoring heterogeneous correlations across different modalities, which will greatly degrade the retrieval performance. In this paper, we propose a novel deep cross-modal hashing method to generate compact hash codes through an end-to-end deep learning architecture, which can effectively capture the intrinsic relationships between various modalities. Our architecture integrates different types of pairwise constraints to encourage the similarities of the hash codes from an intra-modal view and an inter-modal view, respectively. Moreover, additional decorrelation constraints are introduced to this architecture, thus enhancing the discriminative ability of each hash bit. Extensive experiments show that our proposed method yields state-of-the-art results on two cross-modal retrieval datasets.

Visual attention has been successfully applied in structural prediction tasks such as visual captioning and question answering. Existing visual attention models are generally spatial, i.e., the attention is modeled as spatial probabilities that re-weight the last conv-layer feature map of a CNN encoding an input image. However, we argue that such spatial attention does not necessarily conform to the attention mechanism - a dynamic feature extractor that combines contextual fixations over time, as CNN features are naturally spatial, channel-wise and multi-layer. In this paper, we introduce a novel convolutional neural network dubbed SCA-CNN that incorporates Spatial and Channel-wise Attentions in a CNN. In the task of image captioning, SCA-CNN dynamically modulates the sentence generation context in multi-layer feature maps, encoding where (i.e., attentive spatial locations at multiple layers) and what (i.e., attentive channels) the visual attention is. We evaluate the proposed SCA-CNN architecture on three benchmark image captioning datasets: Flickr8K, Flickr30K, and MSCOCO. It is consistently observed that SCA-CNN significantly outperforms state-of-the-art visual attention-based image captioning methods.

Recent research endeavors have shown the potential of using feed-forward convolutional neural networks to accomplish fast style transfer for images. In this work, we take one step further to explore the possibility of exploiting a feed-forward network to perform style transfer for videos and simultaneously maintain temporal consistency among stylized video frames. Our feed-forward network is trained by enforcing the outputs of consecutive frames to be both well stylized and temporally consistent. More specifically, a hybrid loss is proposed to capitalize on the content information of input frames, the style information of a given style image, and the temporal information of consecutive frames. To calculate the temporal loss during the training stage, a novel two-frame synergic training mechanism is proposed. Compared with directly applying an existing image style transfer method to videos, our proposed method employs the trained network to yield temporally consistent stylized videos which are much more visually pleasant. In contrast to the prior video style transfer method which relies on time-consuming optimization on the fly, our method runs in real time while generating competitive visual results.

In this work we study the task of image annotation, of which the goal is to describe an image using a few tags. Instead of predicting the full list of tags, here we target for providing a short list of tags under a limited number (e.g., 3), to cover as much information as possible of the image. The tags in such a short list should be representative and diverse. It means they are required to be not only corresponding to the contents of the image, but also be different to each other. To this end, we treat the image annotation as a subset selection problem based on the conditional determinantal point process (DPP) model, which formulates the representation and diversity jointly. We further explore the semantic hierarchy and synonyms among the candidate tags, and require that two tags in a semantic hierarchy or in a pair of synonyms should not be selected simultaneously. This requirement is then embedded into the sampling algorithm according to the learned conditional DPP model. Besides, we find that traditional metrics for image annotation (e.g., precision, recall and F1 score) only consider the representation, but ignore the diversity. Thus we propose new metrics to evaluate the quality of the selected subset (i.e., the tag list), based on the semantic hierarchy and synonyms. Human study through Amazon Mechanical Turk verifies that the proposed metrics are more close to the humans judgment than traditional metrics. Experiments on two benchmark datasets show that the proposed method can produce more representative and diverse tags, compared with existing image annotation methods.

Deep Convolutional Neural Networks (CNNs) achieve substantial improvements in face detection in the wild. Classical CNN-based face detection methods simply stack successive layers of filters where an input sample should pass through all layers before reaching a face/non-face decision. Inspired by the fact that for face detection, filters in deeper layers can discriminate between difficult face/non-face samples while those in shallower layers can efficiently reject simple non-face samples, we propose Inside Cascaded Structure that introduces face/non-face classifiers at different layers within the same CNN. In the training phase, we propose data routing mechanism which enables different layers to be trained by different types of samples, and thus deeper layers can focus on handling more difficult samples compared with traditional architecture. In addition, we introduce a two-stream contextual CNN architecture that leverages body part information adaptively to enhance face detection. Extensive experiments on the challenging FD-DB and WIDER FACE benchmarks demonstrate that our method achieves competitive accuracy to the state-of-the-art techniques while keeps real time performance.

In this paper, we propose a fast Gauss-Seidel Operator Splitting (GSOS) algorithm for addressing multi-term nonsmooth convex composite optimization, which has wide applications in machine learning, signal processing and statistics. The proposed GSOS algorithm inherits the advantage of the Gauss-Seidel technique to accelerate the optimization procedure, and leverages the operator splitting technique to reduce the computational complexity. In addition, we develop a new technique to establish the global convergence of the GSOS algorithm. To be specific, we first reformulate the iterations of GSOS as a two-step iterations algorithm by employing the tool of operator optimization theory. Subsequently, we establish the convergence of GSOS based on the two-step iterations algorithm reformulation. At last, we apply the proposed GSOS algorithm to solve overlapping group Lasso and graph-guided fused Lasso problems. Numerical experiments show that our proposed GSOS algorithm is superior to the state-of-the-art algorithms in terms of both efficiency and effectiveness.

Sparse support vector machine (SVM) is a popular classification technique that can simultaneously learn a small set of the most interpretable features and identify the support vectors. It has achieved great successes in many real-world applications. However, for large-scale problems involving a huge number of samples and extremely high-dimensional features, solving sparse SVMs remains challenging. By noting that sparse SVMs induce sparsities in both feature and sample spaces, we propose a novel approach, which is based on accurate estimations of the primal and dual optima of sparse SVMs, to simultaneously identify the features and samples that are guaranteed to be irrelevant to the outputs. Thus, we can remove the identified inactive samples and features from the training phase, leading to substantial savings in both the memory usage and computational cost without sacrificing accuracy. To the best of our knowledge, the proposed method is the first static feature and sample reduction method for sparse SVMs. Experiments on both synthetic and real datasets (e.g., the kddb dataset with about 20 million samples and 30 million features) demonstrate that our approach significantly outperforms state-of-the-art methods and the speedup gained by our approach can be orders of magnitude.

Domain adaptation problems arise in a variety of applications, where a training dataset from the source domain and a test dataset from the target domain typically follow different distributions. The primary difficulty in designing effective learning models to solve such problems lies in how to bridge the gap between the source and target distributions. In this paper, we provide comprehensive analysis of feature learning algorithms used in conjunction with linear classifiers for domain adaptation. Our analysis shows that in order to achieve good adaptation performance, the second moments of the source domain distribution and target domain distribution should be similar. Based on our new analysis, a novel extremely easy feature learning algorithm for domain adaptation is proposed. Furthermore, our algorithm is extended by leveraging multiple layers, leading to another feature learning algorithm. We evaluate the effectiveness of the proposed algorithms in terms of domain adaptation tasks on Amazon review and spam datasets from the ECML/PKDD 2006 discovery challenge.

In this paper, we focus on improving Event Extraction (EE) by incorporating visual knowledge with words and phrases from text documents. We first discover visual patterns from large-scale text-image pairs in a weakly-supervised manner and then propose a multimodal event extraction algorithm where the event extractor is jointly trained with textual features and visual patterns. Extensive experimental results on benchmark data sets demonstrate that the proposed multimodal EE method can achieve significantly better performance on event extraction: absolute 7.1% F-score gain on event trigger labeling and 8.5% F-score gain on event argument labeling.

In this paper, we extend the geometric descent method recently proposed by Bubeck, Lee and Singh to tackle nonsmooth and strongly convex composite problems. We prove that our proposed algorithm, dubbed geometric proximal gradient method (GeoPG), converges with a linear rate $(1-1/\sqrt{\kappa})$ and thus achieves the optimal rate among first-order methods, where $\kappa$ is the condition number of the problem. Numerical results on linear regression and logistic regression with elastic net regularization show that GeoPG compares favorably with Nesterov's accelerated proximal gradient method, especially when the problem is ill-conditioned.

Low-rank matrix approximation (LRMA) methods have achieved excellent accuracy among today's collaborative filtering (CF) methods. In existing LRMA methods, the rank of user/item feature matrices is typically fixed, i.e., the same rank is adopted to describe all users/items. However, our studies show that submatrices with different ranks could coexist in the same user-item rating matrix, so that approximations with fixed ranks cannot perfectly describe the internal structures of the rating matrix, therefore leading to inferior recommendation accuracy. In this paper, a mixture-rank matrix approximation (MRMA) method is proposed, in which user-item ratings can be characterized by a mixture of LRMA models with different ranks. Meanwhile, a learning algorithm capitalizing on iterated condition modes is proposed to tackle the non-convex optimization problem pertaining to MRMA. Experimental studies on MovieLens and Netflix datasets demonstrate that MRMA can outperform six state-of-the-art LRMA-based CF methods in terms of recommendation accuracy.

Multimedia content is dominating today's Web information. The nature of multimedia user-item interactions is 1/0 binary implicit feedback (e.g., photo likes, video views, song downloads, etc.), which can be collected at a larger scale with a much lower cost than explicit feedback (e.g., product ratings). However, the majority of existing collaborative filtering (CF) systems are not well-designed for multimedia recommendation, since they ignore the implicitness in users' interactions with multimedia content. We argue that, in multimedia recommendation, there exists item- and component-level implicitness which blurs the underlying users' preferences. The item-level implicitness means that users' preferences on items (e.g. photos, videos, songs, etc.) are unknown, while the component-level implicitness means that inside each item users' preferences on different components (e.g. regions in an image, frames of a video, etc.) are unknown. For example, a 'view'' on a video does not provide any specific information about how the user likes the video (i.e.item-level) and which parts of the video the user is interested in (i.e.component-level). In this paper, we introduce a novel attention mechanism in CF to address the challenging item- and component-level implicit feedback in multimedia recommendation, dubbed Attentive Collaborative Filtering (ACF). Specifically, our attention model is a neural network that consists of two attention modules: the component-level attention module, starting from any content feature extraction network (e.g. CNN for images/videos), which learns to select informative components of multimedia items, and the item-level attention module, which learns to score the item preferences. ACF can be seamlessly incorporated into classic CF models with implicit feedback, such as BPR and SVD++, and efficiently trained using SGD. Through extensive experiments on two real-world multimedia Web services: Vine and Pinterest, we show that ACF significantly outperforms state-of-the-art CF methods.

This paper proposes a generic formulation that significantly expedites the training and deployment of image classification models, particularly under the scenarios of many image categories and high feature dimensions. As the core idea, our method represents both the images and learned classifiers using binary hash codes, which are simultaneously learned from the training data. Classifying an image thereby reduces to retrieving its nearest class codes in the Hamming space. Specifically, we formulate multiclass image classification as an optimization problem over binary variables. The optimization alternatingly proceeds over the binary classifiers and image hash codes. Profiting from the special property of binary codes, we show that the sub-problems can be efficiently solved through either a binary quadratic program (BQP) or a linear program. In particular, for attacking the BQP problem, we propose a novel bit-flipping procedure which enjoys high efficacy and a local optimality guarantee. Our formulation supports a large family of empirical loss functions and is, in specific, instantiated by exponential and linear losses. Comprehensive evaluations are conducted on several representative image benchmarks. The experiments consistently exhibit reduced computational and memory complexities of model training and deployment, without sacrificing classification accuracy.

In this paper, we present a label transfer model from texts to images for image classification tasks. The problem of image classification is often much more challenging than text classification. On one hand, labeled text data is more widely available than the labeled images for classification tasks. On the other hand, text data tends to have natural semantic interpretability, and they are often more directly related to class labels. On the contrary, the image features are not directly related to concepts inherent in class labels. One of our goals in this paper is to develop a model for revealing the functional relationships between text and image features as to directly transfer intermodal and intramodal labels to annotate the images. This is implemented by learning a transfer function as a bridge to propagate the labels between two multimodal spaces. However, the intermodal label transfers could be undermined by blindly transferring the labels of noisy texts to annotate images. To mitigate this problem, we present an intramodal label transfer process, which complements the intermodal label transfer by transferring the image labels instead when relevant text is absent from the source corpus. In addition, we generalize the inter-modal label transfer to zero-shot learning scenario where there are only text examples available to label unseen classes of images without any positive image examples. We evaluate our algorithm on an image classification task and show the effectiveness with respect to the other compared algorithms.

Content-based image retrieval (CBIR) shows great significance in clinical decision-making, which explores the visual content of medical images rather than keywords, tags, or descriptions. It provides doctors an image-guided approach to explore relevant cases that could offer doctors instructive reference. Mammogram screening has been known to be widely used in the early stage diagnosis of breast cancer and could reduce its morbidity and mortality. In this paper, we aim to develop a scalable CBIR method for a large repository of mammogram. To this end, we extend the original Anchor Graph Hashing (AGH) and propose a new unsupervised hashing algorithm, named as composite AGH with iterative quantization (C-AGH-ITQ), which compresses mammographic regions of interest (ROIs) into compact binary codes and enables real-time searching in Hamming space. Multimodal features and different distance metrics are integrated, performing upon a composite Anchor Graph. To improve the effectiveness of the hash code, quantization error is further iteratively minimized by introducing an orthogonal rotation matrix. We evaluate the presented C-AGH-ITQ algorithm on a data set of 11 533 mammographic ROIs obtained from the Digital Database for Screening Mammography. Our method obtains more than 84% retrieval precision and 93% classification accuracy (using$k$NN prediction), which demonstrates that hash codes produced by C-AGH-ITQ well capture the visual similarities between mammographic images. In addition, since C-AGH-ITQ ensures linear complexity of the training procedure and constant time for query, our system is readily applicable to large-scale mammogram databases and has the potential to provide abundant clinical cases as reference.

How can we find a general way to choose the most suitable samples for training a classifier? Even with very limited prior information? Active learning, which can be regarded as an iterative optimization procedure, plays a key role to construct a refined training set to improve the classification performance in a variety of applications, such as text analysis, image recognition, social network modeling, etc. Although combining representativeness and informativeness of samples has been proven promising for active sampling, state-of-the-art methods perform well under certain data structures. Then can we find a way to fuse the two active sampling criteria without any assumption on data? This paper proposes a general active learning framework that effectively fuses the two criteria. Inspired by a two-sample discrepancy problem, triple measures are elaborately designed to guarantee that the query samples not only possess the representativeness of the unlabeled data but also reveal the diversity of the labeled data. Any appropriate similarity measure can be employed to construct the triple measures. Meanwhile, an uncertain measure is leveraged to generate the informativeness criterion, which can be carried out in different ways. Rooted in this framework, a practical active learning algorithm is proposed, which exploits a radial basis function together with the estimated probabilities to construct the triple measures and a modified best-versus-second-best strategy to construct the uncertain measure, respectively. Experimental results on benchmark datasets demonstrate that our algorithm consistently achieves superior performance over the state-of-the-art active learning algorithms.

Recent emergence of low-cost and easy-operating depth cameras has reinvigorated the research in skeleton-based human action recognition. However, most existing approaches overlook the intrinsic interdependencies between skeleton joints and action classes, thus suffering from unsatisfactory recognition performance. In this paper, a novel latent max-margin multitask learning model is proposed for 3-D action recognition. Specifically, we exploit skelets as the mid-level granularity of joints to describe actions. We then apply the learning model to capture the correlations between the latent skelets and action classes each of which accounts for a task. By leveraging structured sparsity inducing regularization, the common information belonging to the same class can be discovered from the latent skelets, while the private information across different classes can also be preserved. The proposed model is evaluated on three challenging action data sets captured by depth cameras. Experimental results show that our model consistently achieves superior performance over recent state-of-the-art approaches.

editorial Free Access Share on Mobile Social Multimedia Analytics in the Big Data Era: An Introduction to the Special Issue Editors: Rongrong Ji Xiamen University, China Xiamen University, ChinaView Profile , Wei Liu Tencent AI Lab, China Tencent AI Lab, ChinaView Profile , Xing Xie Microsoft Research Asia, China Microsoft Research Asia, ChinaView Profile , Yiqiang Chen Chinese Academy of Science, China Chinese Academy of Science, ChinaView Profile , Jiebo Luo University of Rochester, United States University of Rochester, United StatesView Profile Authors Info & Claims ACM Transactions on Intelligent Systems and TechnologyVolume 8Issue 3May 2017 Article No.: 34pp 1–3https://doi.org/10.1145/3040934Published:14 April 2017Publication History 2citation224DownloadsMetricsTotal Citations2Total Downloads224Last 12 Months11Last 6 weeks2 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my AlertsNew Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteeReaderPDF

Stochastic gradient descent (SGD) holds as a classical method to build large scale machine learning models over big data. A stochastic gradient is typically calculated from a limited number of samples (known as mini-batch), which potentially incurs a high variance and causes the estimated parameters to bounce around the optimal solution. To improve the stability of stochastic gradient, recent years have witnessed the proposal of several semi-stochastic gradient descent algorithms, which distinguish themselves from standard SGD by incorporating global information into gradient computation. In this paper, we contribute a novel stratified semi-stochastic gradient descent (S3GD) algorithm to this nascent research area, accelerating the optimization of a large family of composite convex functions. Though theoretically converging faster, prior semi-stochastic algorithms are found to suffer from high iteration complexity, which makes them even slower than SGD in practice on many datasets. In our proposed S3GD, the semi-stochastic gradient is calculated based on efficient manifold propagation, which can be numerically accomplished by sparse matrix multiplications. This way S3GD is able to generate a highly-accurate estimate of the exact gradient from each mini-batch with largely-reduced computational complexity. Theoretic analysis reveals that the proposed S3GD elegantly balances the geometric algorithmic convergence rate against the space and time complexities during the optimization. The efficacy of S3GD is also experimentally corroborated on several large-scale benchmark datasets.

Learning to hash has attracted broad research interests in recent computer vision and machine learning studies, due to its ability to accomplish efficient approximate nearest neighbor search. However, the closely related task, maximum inner product search (MIPS), has rarely been studied in this literature. To facilitate the MIPS study, in this paper, we introduce a general binary coding framework based on asymmetric hash functions, named asymmetric inner-product binary coding (AIBC). In particular, AIBC learns two different hash functions, which can reveal the inner products between original data vectors by the generated binary vectors. Although conceptually simple, the associated optimization is very challenging due to the highly nonsmooth nature of the objective that involves sign functions. We tackle the nonsmooth optimization in an alternating manner, by which each single coding function is optimized in an efficient discrete manner. We also simplify the objective by discarding the quadratic regularization term which significantly boosts the learning efficiency. Both problems are optimized in an effective discrete way without continuous relaxations, which produces high-quality hash codes. In addition, we extend the AIBC approach to the supervised hashing scenario, where the inner products of learned binary codes are forced to fit the supervised similarities. Extensive experiments on several benchmark image retrieval databases validate the superiority of the AIBC approaches over many recently proposed hashing algorithms.

As the prosperity of low-cost and easy-operating depth cameras, skeleton-based human action recognition has been extensively studied recently. However, most of the existing methods partially consider that all 3D joints of a human skeleton are identical. Actually, these 3D joints exhibit diverse responses to different action classes, and some joint configurations are more discriminative to distinguish a certain action. In this paper, we propose a discriminative multi-instance multitask learning (MIMTL) framework to discover the intrinsic relationship between joint configurations and action classes. First, a set of discriminative and informative joint configurations for the corresponding action class is captured in multi-instance learning model by regarding the action and the joint configurations as a bag and its instances, respectively. Then, a multitask learning model with group structure constraints is exploited to further reveal the intrinsic relationship between the joint configurations and different action classes. We conduct extensive evaluations of MIMTL using three benchmark 3D action recognition datasets. Experimental results show that our proposed MIMTL framework performs favorably compared with several state-of-the-art approaches.

How to propagate label information from labeled examples to unlabeled examples over a graph has been intensively studied for a long time. Existing graph-based propagation algorithms usually treat unlabeled examples equally, and transmit seed labels to the unlabeled examples that are connected to the labeled examples in a neighborhood graph. However, such a popular propagation scheme is very likely to yield inaccurate propagation, because it falls short of tackling ambiguous but critical data points (e.g., outliers). To this end, this paper treats the unlabeled examples in different levels of difficulties by assessing their reliability and discriminability, and explicitly optimizes the propagation quality by manipulating the propagation sequence to move from simple to difficult examples. In particular, we propose a novel iterative label propagation algorithm in which each propagation alternates between two paradigms, teaching-to-learn and learning-to-teach (TLLT). In the teaching-to-learn step, the learner conducts the propagation on the simplest unlabeled examples designated by the teacher. In the learning-to-teach step, the teacher incorporates the learner's feedback to adjust the choice of the subsequent simplest examples. The proposed TLLT strategy critically improves the accuracy of label propagation, making our algorithm substantially robust to the values of tuning parameters, such as the Gaussian kernel width used in graph construction. The merits of our algorithm are theoretically justified and empirically demonstrated through experiments performed on both synthetic and real-world data sets.

Multi-label propagation aims to transmit the multi-label information from labeled examples to unlabeled examples based on a weighted graph. Existing methods ignore the specific propagation difficulty of different unlabeled examples and conduct the propagationin an imperfect sequence, leading to the error-prone classification of some difficult examples with uncertain labels. To address this problem, this paper associates each possible label with a "teacher", and proposesa "Multi-Label Teaching-to-Learn and Learning-to-Teach" (ML-TLLT) algorithm, so that the entire propagationprocess is guided by the teachers and manipulated from simple examples to more difficult ones. In the teaching-to-learn step, the teachers select the simplest examples for the current propagation by investigating both the definitiveness of each possible label of the unlabeled examples, and the dependencies between labels revealed by the labeled examples. In the learning-to-teach step, the teachers reversely learn from the learner’s feedback to properly select the simplest examples for the next propagation. Thorough empirical studies show that due to the optimized propagation sequence designed by the teachers, ML-TLLT yields generally better performance than seven state-of-the-art methods on the typical multi-label benchmark datasets.

In the past decades, Spectral Clustering (SC) has become one of the most effective clustering approaches. Although it has been widely used, one significant drawback of SC is its expensive computation cost. Many efforts have been devoted to accelerating SC algorithms and promising results have been achieved. However, most of the existing algorithms rely on the assumption that data can be stored in the computer memory. When data cannot fit in the memory, these algorithms will suffer severe performance degradations. In order to overcome this issue, we propose a novel sequential SC algorithm for tackling large-scale clustering with limited computational resources, \textit{e.g.}, memory. We begin with investigating an effective way of approximating the graph affinity matrix via leveraging a bipartite graph. Then we choose a smart graph construction and optimization strategy to avoid random access to data. These efforts lead to an efficient SC algorithm whose memory usage is independent of the number of input data points. Extensive experiments carried out on large datasets demonstrate that the proposed sequential SC algorithm is up to a thousand times faster than the state-of-the-arts.

Binary code learning, a.k.a., hashing, has been recently popular due to its high efficiency in large-scale similarity search and recognition. It typically maps high-dimensional data points to binary codes, where data similarity can be efficiently computed via rapid Hamming distance. Most existing unsupervised hashing schemes pursue binary codes by reducing the quantization error from an original real-valued data space to a resulting Hamming space. On the other hand, most existing supervised hashing schemes constrain binary code learning to correlate with pairwise similarity labels. However, few methods consider ordinal relations in the binary code learning process, which serve as a very significant cue to learn the optimal binary codes for similarity search. In this paper, we propose a novel hashing scheme, dubbed Ordinal Embedding Hashing (OEH), which embeds given ordinal relations among data points to learn the ranking-preserving binary codes. The core idea is to construct a directed unweighted graph to capture the ordinal relations, and then train the hash functions using this ordinal graph to preserve the permutation relations in the Hamming space. To learn such hash functions effectively, we further relax the discrete constraints and design a stochastic gradient decent algorithm to obtain the optimal solution. Experimental results on two large-scale benchmark datasets demonstrate that the proposed OEH method can achieve superior performance over the state-of-the-arts approaches.At last, the evaluation on query by humming dataset demonstrates the OEH also has good performance for music retrieval by using user's humming or singing.

This paper studies the Tensor Robust Principal Component (TRPCA) problem which extends the known Robust PCA [4] to the tensor case. Our model is based on a new tensor Singular Value Decomposition (t-SVD) [14] and its induced tensor tubal rank and tensor nuclear norm. Consider that we have a 3-way tensor X ε Rn1×n2×n3such that X = L0+ S0, where L0has low tubal rank andS0 issparse. Is that possible to recover both components? In this work, we prove that under certain suitable assumptions, we can recover both the low-rank and the sparse components exactly by simply solving a convex program whose objective is a weighted combination of the tensor nuclear norm and the l1-norm, i.e., min L, E ||L||* + λ||ε||1, s.t. X = L + ε, where λ = 1/√max(n1, n2)n3. Interestingly, TRPCA involves RPCA as a special case when n3= 1 and thus it is a simple and elegant tensor extension of RPCA. Also numerical experiments verify our theory and the application for the image denoising demonstrates the effectiveness of our method.

We consider a wide spectrum of regularized stochastic minimization problems, where the regularization term is composite with a linear function. Examples of this formulation include graph-guided regularized minimization, generalized Lasso and a class of ℓ1 regularized problems. The computational challenge is that the closed-form solution of the proximal mapping associated with the regularization term is not available due to the imposed linear composition. Fortunately, the structure of the regularization term allows us to reformulate it as a new convex-concave saddle point problem which can be solved using the Primal-Dual Hybrid Gradient (PDHG) approach. However, this approach may be inefficient in realistic applications as computing the full gradient of the expected objective function could be very expensive when the number of input data samples is considerably large. To address this issue, we propose a Stochastic PDHG (SPDHG) algorithm with either uniformly or non-uniformly averaged iterates. Through uniformly averaged iterates, the SPDHG algorithm converges in expectation withrate for general convex objectives and O(log (t)/t) rate for strongly convex objectives, respectively. While with non-uniformly averaged iterates, the SPDHG algorithm is expected to converge with O(1/t) rate for strongly convex objectives. Numerical experiments on different genres of datasets demonstrate that our proposed algorithm outperforms other competing algorithms.

We address the efficiency problem of Collaborative Filtering (CF) by hashing users and items as latent vectors in the form of binary codes, so that user-item affinity can be efficiently calculated in a Hamming space. However, existing hashing methods for CF employ binary code learning procedures that most suffer from the challenging discrete constraints. Hence, those methods generally adopt a two-stage learning scheme composed of relaxed optimization via discarding the discrete constraints, followed by binary quantization. We argue that such a scheme will result in a large quantization loss, which especially compromises the performance of large-scale CF that resorts to longer binary codes. In this paper, we propose a principled CF hashing framework called Discrete Collaborative Filtering (DCF), which directly tackles the challenging discrete optimization that should have been treated adequately in hashing. The formulation of DCF has two advantages: 1) the Hamming similarity induced loss that preserves the intrinsic user-item similarity, and 2) the balanced and uncorrelated code constraints that yield compact yet informative binary codes. We devise a computationally efficient algorithm with a rigorous convergence proof of DCF. Through extensive experiments on several real-world benchmarks, we show that DCF consistently outperforms state-of-the-art CF hashing techniques, e.g, though using only 8 bits, DCF is even significantly better than other methods using 128 bits.

The explosive growth in Big Data has attracted much attention in designing efficient indexing and search methods recently. In many critical applications such as large-scale search and pattern matching, finding the nearest neighbors to a query is a fundamental research problem. However, the straightforward solution using exhaustive comparison is infeasible due to the prohibitive computational complexity and memory requirement. In response, approximate nearest neighbor (ANN) search based on hashing techniques has become popular due to its promising performance in both efficiency and accuracy. Prior randomized hashing methods, e.g., locality-sensitive hashing (LSH), explore data-independent hash functions with random projections or permutations. Although having elegant theoretic guarantees on the search quality in certain metric spaces, performance of randomized hashing has been shown insufficient in many real-world applications. As a remedy, new approaches incorporating data-driven learning methods in development of advanced hash functions have emerged. Such learning-to-hash methods exploit information such as data distributions or class labels when optimizing the hash codes or functions. Importantly, the learned hash codes are able to preserve the proximity of neighboring data in the original feature spaces in the hash code spaces. The goal of this paper is to provide readers with systematic understanding of insights, pros, and cons of the emerging techniques. We provide a comprehensive survey of the learning-to-hash framework and representative techniques of various types, including unsupervised, semisupervised, and supervised. In addition, we also summarize recent hashing approaches utilizing the deep learning models. Finally, we discuss the future direction and trends of research in this area.

Semi-supervised image classification aims to classify a large quantity of unlabeled images by typically harnessing scarce labeled images. Existing semi-supervised methods often suffer from inadequate classification accuracy when encountering difficult yet critical images, such as outliers, because they treat all unlabeled images equally and conduct classifications in an imperfectly ordered sequence. In this paper, we employ the curriculum learning methodology by investigating the difficulty of classifying every unlabeled image. The reliability and the discriminability of these unlabeled images are particularly investigated for evaluating their difficulty. As a result, an optimized image sequence is generated during the iterative propagations, and the unlabeled images are logically classified from simple to difficult. Furthermore, since images are usually characterized by multiple visual feature descriptors, we associate each kind of features with a teacher, and design a multi-modal curriculum learning (MMCL) strategy to integrate the information from different feature modalities. In each propagation, each teacher analyzes the difficulties of the currently unlabeled images from its own modality viewpoint. A consensus is subsequently reached among all the teachers, determining the currently simplest images (i.e., a curriculum), which are to be reliably classified by the multi-modal learner. This well-organized propagation process leveraging multiple teachers and one learner enables our MMCL to outperform five state-of-the-art methods on eight popular image data sets.

Capabilities of inference and prediction are the significant components of visual systems. Visual path prediction is an important and challenging task among them, with the goal to infer the future path of a visual object in a static scene. This task is complicated as it needs high-level semantic understandings of both the scenes and underlying motion patterns in video sequences. In practice, cluttered situations have also raised higher demands on the effectiveness and robustness of models. Motivated by these observations, we propose a deep learning framework, which simultaneously performs deep feature learning for visual representation in conjunction with spatiotemporal context modeling. After that, a unified path-planning scheme is proposed to make accurate path prediction based on the analytic results returned by the deep context models. The highly effective visual representation and deep context models ensure that our framework makes a deep semantic understanding of the scenes and motion patterns, consequently improving the performance on visual path prediction task. In experiments, we extensively evaluate the model's performance by constructing two large benchmark datasets from the adaptation of video tracking datasets. The qualitative and quantitative experimental results show that our approach outperforms the state-of-the-art approaches and owns a better generalization capability.

Cross-modal retrieval has attracted much attention in recent years due to its widespread applications. In this area, how to capture and correlate heterogeneous features originating from different modalities remains a challenge. However, most existing methods dealing with cross-modal learning only focus on learning relevant features shared by two distinct feature spaces, therefore overlooking discriminative feature information of them. To remedy this issue and explicitly capture discriminative feature information, we propose a novel cross-modal retrieval approach based on discriminative dictionary learning that is augmented with common label alignment. Concretely, a discriminative dictionary is first learned to account for each modality, which boosts not only the discriminating capability of intra-modality data from different classes but also the relevance of inter-modality data in the same class. Subsequently, all the resulting sparse codes are simultaneously mapped to a common label space, where the cross-modal data samples are characterized and associated. Also in the label space, the discriminativeness and relevance of the considered cross-modal data can be further strengthened by enforcing a common label alignment. Finally, cross-modal retrieval is performed over the common label space. Experiments conducted on two public cross-modal datasets show that the proposed approach outperforms several state-of-the-art methods in term of retrieval accuracy.

Existing video search engines return a ranked list of videos for each user query, which is not convenient for browsing the results of query topics that have multiple facets, such as the “early life,” “personal life,” and “presidency” of a query “Barack Obama.” Organizing video search results into semantically structured hierarchies with nodes covering different topic facets can significantly improve the browsing efficiency for such queries. In this paper, we introduce a hierarchical visualization approach for video search result browsing, which can help users quickly understand the multiple facets of a query topic in a very well-organized manner. Given a query, our approach starts from the hierarchy of its textual descriptions normally available on Wikipedia and then adjusts the hierarchical structure by analyzing the video information to reflect the topic structure of the search result. After that, a simple optimization problem is formulated to perform the video-to-node association considering three important criteria. Furthermore, additional topic facets are mined to complement the contents of the existing semantic hierarchies. A large YouTube video dataset is constructed to evaluate our approach both quantitatively and qualitatively. A demo system is also developed for users to interact with the proposed browsing approach.

Multiple-Instance (MI) learning is an important supervised learning technique which deals with collections of instances called bags. While existing research in MI learning mainly focused on classification, in this paper we propose a new approach for MI retrieval to enable effective similarity retrieval of bags of instances, where training data is presented in the form of similar and dissimilar bag pairs. An embedded scheme is devised as encoding each bag into a single bag feature vector by exploiting a similarity-based transformation. In this way, the original MI problem is converted into a single-instance version. Furthermore, we develop a principled approach for optimizing bag features specific to similarity retrieval through leveraging pairwise label information at the bag level. The experimental results demonstrate the effectiveness of the proposed approach in comparison with the alternatives for MI retrieval.

Metric learning has become a widespreadly used tool in machine learning. To reduce expensive costs brought in by increasing dimensionality, low-rank metric learning arises as it can be more economical in storage and computation. However, existing low-rank metric learning algorithms usually adopt nonconvex objectives, and are hence sensitive to the choice of a heuristic low-rank basis. In this paper, we propose a novel low-rank metric learning algorithm to yield bilinear similarity functions. This algorithm scales linearly with input dimensionality in both space and time, therefore applicable to high-dimensional data domains. A convex objective free of heuristics is formulated by leveraging trace norm regularization to promote low-rankness. Crucially, we prove that all globally optimal metric solutions must retain a certain low-rank structure, which enables our algorithm to decompose the high-dimensional learning task into two steps: an SVD-based projection and a metric learning problem with reduced dimensionality. The latter step can be tackled efficiently through employing a linearized Alternating Direction Method of Multipliers. The efficacy of the proposed algorithm is demonstrated through experiments performed on four benchmark datasets with tens of thousands of dimensions.

Saliency propagation has been widely adopted for identifying the most attractive object in an image. The propagation sequence generated by existing saliency detection methods is governed by the spatial relationships of image regions, i.e., the saliency value is transmitted between two adjacent regions. However, for the inhomogeneous difficult adjacent regions, such a sequence may incur wrong propagations. In this paper, we attempt to manipulate the propagation sequence for optimizing the propagation quality. Intuitively, we postpone the propagations to difficult regions and meanwhile advance the propagations to less ambiguous simple regions. Inspired by the theoretical results in educational psychology, a novel propagation algorithm employing the teaching-to-learn and learning-to-teach strategies is proposed to explicitly improve the propagation quality. In the teaching-to-learn step, a teacher is designed to arrange the regions from simple to difficult and then assign the simplest regions to the learner. In the learning-to-teach step, the learner delivers its learning confidence to the teacher to assist the teacher to choose the subsequent simple regions. Due to the interactions between the teacher and learner, the uncertainty of original difficult regions is gradually reduced, yielding manifest salient objects with optimized background suppression. Extensive experimental results on benchmark saliency datasets demonstrate the superiority of the proposed algorithm over twelve representative saliency detectors.

Nowadays, detecting objects in 3D scenes like point clouds has become an emerging challenge with various applications. However, it retains as an open problem due to the deficiency of labeling 3D training data. To deploy an accurate detection algorithm typically resorts to investigating both RGB and depth modalities, which have distinct statistics while correlated with each other. Previous research mainly focus on detecting objects using only one modality, which ignores exploiting the cross-modality cues. In this work, we propose a cross-modality deep learning framework based on deep Boltzmann Machines for 3D Scenes object detection. In particular, we demonstrate that by learning cross-modality feature from RGBD data, it is possible to capture their joint information to reinforce detector trainings in individual modalities. In particular, we slide a 3D detection window in the 3D point cloud to match the exemplar shape, which the lack of training data in 3D domain is conquered via (1) We collect 3D CAD models and 2D positive samples from Internet. (2) adopt pretrained R-CNNs [2] to extract raw feature from both RGB and Depth domains. Experiments on RMRC dataset demonstrate that the bimodal based deep feature learning framework helps 3D scene object detection.

Exploring image structure is a long-standing yet important research subject in the computer vision community. In this paper, we focus on understanding image structure inspired by the “simple-to-complex” biological evidence. A hierarchical shape parsing strategy is proposed to partition and organize image components into a hierarchical structure in the scale space. To improve the robustness and flexibility of image representation, we further bundle the image appearances into hierarchical parsing trees. Image descriptions are subsequently constructed by performing a structural pooling, facilitating efficient matching between the parsing trees. We leverage the proposed hierarchical shape parsing to study two exemplar applications including edge scale refinement and unsupervised “objectness” detection. We show competitive parsing performance comparing to the state-of-the-arts in above scenarios with far less proposals, which thus demonstrates the advantage of the proposed parsing scheme.

Recently, learning based hashing techniques have attracted broad research interests because they can support efficient storage and retrieval for high-dimensional data such as images, videos, documents, etc. However, a major difficulty of learning to hash lies in handling the discrete constraints imposed on the pursued hash codes, which typically makes hash optimizations very challenging (NP-hard in general). In this work, we propose a new supervised hashing framework, where the learning objective is to generate the optimal binary hash codes for linear classification. By introducing an auxiliary variable, we reformulate the objective such that it can be solved substantially efficiently by employing a regularization algorithm. One of the key steps in this algorithm is to solve a regularization sub-problem associated with the NP-hard binary optimization. We show that the sub-problem admits an analytical solution via cyclic coordinate descent. As such, a high-quality discrete solution can eventually be obtained in an efficient computing manner, therefore enabling to tackle massive datasets. We evaluate the proposed approach, dubbed Supervised Discrete Hashing (SDH), on four large image datasets and demonstrate its superiority to the state-of-the-art hashing methods in large-scale image retrieval.

This paper focuses on the problem of hyper-graph matching, by accounting for both unary and higher-order affinity terms. Our method is in line with the linear approximate framework while the problem is iteratively solved in discrete space. It is empirically found more efficient than many extant continuous methods. Moreover, it avoids unknown accuracy loss by heuristic rounding step from the continuous approaches. Under weak assumptions, we prove the iterative discrete gradient assignment in general will trap into a degenerating case - an m-circle solution path where m is the order of the problem. A tailored adaptive relaxation mechanism is devised to detect the degenerating case and makes the algorithm converge to a fixed point in discrete space. Evaluations on both synthetic and real-world data corroborate the efficiency of our method.

In recent years, hashing techniques are becoming overwhelmingly popular for their high efficiency in handling large-scale computer vision applications. It has been shown that hashing techniques which leverage supervised information can significantly enhance performance, and thus greatly benefit visual search tasks. Typically, a modern hashing method uses a set of hash functions to compress data samples into compact binary codes. However, few methods have developed hash functions to optimize the precision at the top of a ranking list based upon Hamming distances. In this paper, we propose a novel supervised hashing approach, namely Rank Preserving Hashing (RPH), to explicitly optimize the precision of Hamming distance ranking towards preserving the supervised rank information. The core idea is to train disciplined hash functions in which the mistakes at the top of a Hamming-distance ranking list are penalized more than those at the bottom. To find such hash functions, we relax the original discrete optimization objective to a continuous surrogate, and then design an online learning algorithm to efficiently optimize the surrogate objective. Empirical studies based upon two benchmark image datasets demonstrate that the proposed hashing approach achieves superior image search accuracy over the state-of-the-art approaches.

Binary coding or hashing techniques are recognized to accomplish efficient near neighbor search, and have thus attracted broad interests in the recent vision and learning studies. However, such studies have rarely been dedicated to Maximum Inner Product Search (MIPS), which plays a critical role in various vision applications. In this paper, we investigate learning binary codes to exclusively handle the MIPS problem. Inspired by the latest advance in asymmetric hashing schemes, we propose an asymmetric binary code learning framework based on inner product fitting. Specifically, two sets of coding functions are learned such that the inner products between their generated binary codes can reveal the inner products between original data vectors. We also propose an alternative simpler objective which maximizes the correlations between the inner products of the produced binary codes and raw data vectors. In both objectives, the binary codes and coding functions are simultaneously learned without continuous relaxations, which is the key to achieving high-quality binary codes. We evaluate the proposed method, dubbed Asymmetric Inner-product Binary Coding (AIBC), relying on the two objectives on several large-scale image datasets. Both of them are superior to the state-of-the-art binary coding and hashing methods in performing MIPS tasks.

In recent years, binary coding techniques are becoming increasingly popular because of their high efficiency in handling large-scale computer vision applications. It has been demonstrated that supervised binary coding techniques that leverage supervised information can significantly enhance the coding quality, and hence greatly benefit visual search tasks. Typically, a modern binary coding method seeks to learn a group of coding functions which compress data samples into binary codes. However, few methods pursued the coding functions such that the precision at the top of a ranking list according to Hamming distances of the generated binary codes is optimized. In this paper, we propose a novel supervised binary coding approach, namely Top Rank Supervised Binary Coding (Top-RSBC), which explicitly focuses on optimizing the precision of top positions in a Hamming-distance ranking list towards preserving the supervision information. The core idea is to train the disciplined coding functions, by which the mistakes at the top of a Hamming-distance ranking list are penalized more than those at the bottom. To solve such coding functions, we relax the original discrete optimization objective with a continuous surrogate, and derive a stochastic gradient descent to optimize the surrogate objective. To further reduce the training time cost, we also design an online learning algorithm to optimize the surrogate objective more efficiently. Empirical studies based upon three benchmark image datasets demonstrate that the proposed binary coding approach achieves superior image search accuracy over the state-of-the-arts.

Semi-supervised clustering leverages side information such as pairwise constraints to guide clustering procedures. Despite promising progress, existing semi-supervised clustering approaches overlook the condition of side information being generated sequentially, which is a natural setting arising in numerous real-world applications such as social network and e-commerce system analysis. Given emerged new constraints, classical semi-supervised clustering algorithms need to re-optimize their objectives over all data samples and constraints in availability, which prevents them from efficiently updating the obtained data partitions. To address this challenge, we propose an efficient dynamic semi-supervised clustering framework that casts the clustering problem into a search problem over a feasible convex set, i.e., a convex hull with its extreme points being an ensemble of m data partitions. According to the principle of ensemble clustering, the optimal partition lies in the convex hull, and can thus be uniquely represented by an m-dimensional probability simplex vector. As such, the dynamic semi-supervised clustering problem is simplified to the problem of updating a probability simplex vector subject to the newly received pairwise constraints. We then develop a computationally efficient updating procedure to update the probability simplex vector in O(m2) time, irrespective of the data size n. Our empirical studies on several real-world benchmark datasets show that the proposed algorithm outperforms the state-of-the-art semi-supervised clustering algorithms with visible performance gain and significantly reduced running time.

In this paper, we propose a novel method for image fusion with a high-resolution panchromatic image and a low-resolution multispectral (Ms) image at the same geographical location. The fusion is formulated as a convex optimization problem which minimizes a linear combination of a least-squares fitting term and a dynamic gradient sparsity regularizer. The former is to preserve accurate spectral information of the Ms image, while the latter is to keep sharp edges of the high-resolution panchromatic image. We further propose to simultaneously register the two images during the fusing process, which is naturally achieved by virtue of the dynamic gradient sparsity property. An efficient algorithm is then devised to solve the optimization problem, accomplishing a linear computational complexity in the size of the output image in each iteration. We compare our method against six state-of-the-art image fusion methods on Ms image data sets from four satellites. Extensive experimental results demonstrate that the proposed method substantially outperforms the others in terms of both spatial and spectral qualities. We also show that our method can provide high-quality products from coarsely registered real-world IKONOS data sets. Finally, a MATLAB implementation is provided to facilitate future research.

Human action recognition in unconstrained videos is a challenging problem with many applications. Most state-of-the-art approaches adopted the well-known bag-of-features representations, generated based on isolated local patches or patch trajectories, where motion patterns, such as object-object and object-background relationships are mostly discarded. In this paper, we propose a simple representation aiming at modeling these motion relationships. We adopt global and local reference points to explicitly characterize motion information, so that the final representation is more robust to camera movements, which widely exist in unconstrained videos. Our approach operates on the top of visual codewords generated on dense local patch trajectories, and therefore, does not require foreground-background separation, which is normally a critical and difficult step in modeling object relationships. Through an extensive set of experimental evaluations, we show that the proposed representation produces a very competitive performance on several challenging benchmark data sets. Further combining it with the standard bag-of-features or Fisher vector representations can lead to substantial improvements.

Conditional random fields (CRFs) are a flexible yet powerful probabilistic approach and have shown advantages for popular applications in various areas, including text analysis, bioinformatics, and computer vision. Traditional CRF models, however, are incapable of selecting relevant features as well as suppressing noise from noisy original features. Moreover, conventional optimization methods often converge slowly in solving the training procedure of CRFs, and will degrade significantly for tasks with a large number of samples and features. In this paper, we propose robust CRFs (RCRFs) to simultaneously select relevant features. An optimal gradient method (OGM) is further designed to train RCRFs efficiently. Specifically, the proposed RCRFs employ the l1 norm of the model parameters to regularize the objective used by traditional CRFs, therefore enabling discovery of the relevant unary features and pairwise features of CRFs. In each iteration of OGM, the gradient direction is determined jointly by the current gradient together with the historical gradients, and the Lipschitz constant is leveraged to specify the proper step size. We show that an OGM can tackle the RCRF model training very efficiently, achieving the optimal convergence rate [Formula: see text] (where k is the number of iterations). This convergence rate is theoretically superior to the convergence rate O(1/k) of previous first-order optimization methods. Extensive experiments performed on three practical image segmentation tasks demonstrate the efficacy of OGM in training our proposed RCRFs.

Image denoising is a fundamental problem in computer vision and image processing that holds considerable practical importance for real-world applications. The traditional patch-based and sparse coding-driven image denoising methods convert 2D image patches into 1D vectors for further processing. Thus, these methods inevitably break down the inherent 2D geometric structure of natural images. To overcome this limitation pertaining to the previous image denoising methods, we propose a 2D image denoising model, namely, the dictionary pair learning (DPL) model, and we design a corresponding algorithm called the DPL on the Grassmann-manifold (DPLG) algorithm. The DPLG algorithm first learns an initial dictionary pair (i.e., the left and right dictionaries) by employing a subspace partition technique on the Grassmann manifold, wherein the refined dictionary pair is obtained through a sub-dictionary pair merging. The DPLG obtains a sparse representation by encoding each image patch only with the selected sub-dictionary pair. The non-zero elements of the sparse representation are further smoothed by the graph Laplacian operator to remove the noise. Consequently, the DPLG algorithm not only preserves the inherent 2D geometric structure of natural images but also performs manifold smoothing in the 2D sparse coding space. We demonstrate that the DPLG algorithm also improves the structural SIMilarity values of the perceptual visual quality for denoised images using the experimental evaluations on the benchmark images and Berkeley segmentation data sets. Moreover, the DPLG also produces the competitive peak signal-to-noise ratio values from popular image denoising algorithms.

Coming with the popularity of multimedia sharing platforms such as Facebook and Flickr, recent years have witnessed an explosive growth of geographical tags on social multimedia content. This trend enables a wide variety of emerging applications, for example, mobile location search, landmark recognition, scene reconstruction, and touristic recommendation, which range from purely research prototype to commercial systems. In this article, we give a comprehensive survey on these applications, covering recent advances in recognition and mining of geographical-aware social multimedia. We review related work in the past decade regarding to location recognition, scene summarization, tourism suggestion, 3D building modeling, mobile visual search and city navigation. At the end, we further discuss potential challenges, future topics, as well as open issues related to geo-social multimedia computing, recognition, mining, and analytics.

Automatic analysis of histopathological images has been widely utilized leveraging computational image-processing methods and modern machine learning techniques. Both computer-aided diagnosis (CAD) and content-based image-retrieval (CBIR) systems have been successfully developed for diagnosis, disease detection, and decision support in this area. Recently, with the ever-increasing amount of annotated medical data, large-scale and data-driven methods have emerged to offer a promise of bridging the semantic gap between images and diagnostic information. In this paper, we focus on developing scalable image-retrieval techniques to cope intelligently with massive histopathological images. Specifically, we present a supervised kernel hashing technique which leverages a small amount of supervised information in learning to compress a 10 000-dimensional image feature vector into only tens of binary bits with the informative signatures preserved. These binary codes are then indexed into a hash table that enables real-time retrieval of images in a large database. Critically, the supervised information is employed to bridge the semantic gap between low-level image features and high-level diagnostic information. We build a scalable image-retrieval framework based on the supervised hashing technique and validate its performance on several thousand histopathological images acquired from breast microscopic tissues. Extensive evaluations are carried out in terms of image classification (i.e., benign versus actionable categorization) and retrieval tests. Our framework achieves about 88.1% classification accuracy as well as promising time efficiency. For example, the framework can execute around 800 queries in only 0.01 s, comparing favorably with other commonly used dimensionality reduction and feature selection methods.

Recently with the explosive growth of visual content on the Internet, large-scale image search has attracted intensive attention. It has been shown that mapping highdimensional image descriptors to compact binary codes can lead to considerable efficiency gains in both storage and similarity computation of images. However, most existing methods still suffer from expensive training devoted to large-scale binary code learning. To address this issue, we propose a sub-selection based matrix manipulation algorithm which can significantly reduce the computational cost of code learning. As case studies, we apply the sub-selection algorithm to two popular quantization techniques PCA Quantization (PCAQ) and Iterative Quantization (ITQ). Crucially, we can justify the resulting sub-selective quantization by proving its theoretic properties. Extensive experiments are carried out on three image benchmarks with up to one million samples, corroborating the efficacy of the sub-selective quantization method in terms of image retrieval.

In this paper, we propose a novel method for image fusion from a high resolution panchromatic image and a low resolution multispectral image at the same geographical location. Different from previous methods, we do not make any assumption about the upsampled multispectral image, but only assume that the fused image after downsampling should be close to the original multispectral image. This is a severely ill-posed problem and a dynamic gradient sparsity penalty is thus proposed for regularization. Incorporating the intra- correlations of different bands, this penalty can effectively exploit the prior information (e.g. sharp boundaries) from the panchromatic image. A new convex optimization algorithm is proposed to efficiently solve this problem. Extensive experiments on four multispectral datasets demonstrate that the proposed method significantly outperforms the state-of-the-arts in terms of both spatial and spectral qualities.

Outliers are pervasive in many computer vision and pattern recognition problems. Automatically eliminating outliers scattering among practical data collections becomes increasingly important, especially for Internet inspired vision applications. In this paper, we propose a novel one-class learning approach which is robust to contamination of input training data and able to discover the outliers that corrupt one class of data source. Our approach works under a fully unsupervised manner, differing from traditional one-class learning supervised by known positive labels. By design, our approach optimizes a kernel-based max-margin objective which jointly learns a large margin one-class classifier and a soft label assignment for inliers and outliers. An alternating optimization algorithm is then designed to iteratively refine the classifier and the labeling, achieving a provably convergent solution in only a few iterations. Extensive experiments conducted on four image datasets in the presence of artificial and real-world outliers demonstrate that the proposed approach is considerably superior to the state-of-the-arts in obliterating outliers from contaminated one class of images, exhibiting strong robustness at a high outlier proportion up to 60%.

Caching query results is an efficient technique for Web search engines. Admission policy can prevent infrequent queries from taking space of more frequent queries in the cache. In this paper we present two novel admission policies tailored for query results cache. These policies are based on query results prefetching information. We also propose a demote operation for the query results cache to improve the cache hit ratio. We then use a trace of over 5 million queries to evaluate our admission policies, as well as traditional policies. Experimental results show that our prefetch-aware admission policies can achieve hit ratios better than state-of-the-art admission policies.

This paper describes and provides an initial solution to a novel video editing task, i.e., video de-fencing. It targets automatic restoration of the video clips that are corrupted by fence-like occlusions during capture. Our key observation lies in the visual parallax between fences and background scenes, which is caused by the fact that the former are typically closer to the camera. Unlike in traditional image inpainting, fence-occluded pixels in the videos tend to appear later in the temporal dimension and are therefore recoverable via optimized pixel selection from relevant frames. To eventually produce fence-free videos, major challenges include cross-frame subpixel image alignment under diverse scene depth, and correct pixel selection that is robust to dominating fence pixels. Several novel tools are developed in this paper, including soft fence detection, weighted truncated optical flow method, and robust temporal median filter. The proposed algorithm is validated on several real-world video clips with fences.

Bag-of-features (BoFs) representation has been extensively applied to deal with various computer vision applications. To extract discriminative and descriptive BoF, one important step is to learn a good dictionary to minimize the quantization loss between local features and codewords. While most existing visual dictionary learning approaches are engaged with unsupervised feature quantization, the latest trend has turned to supervised learning by harnessing the semantic labels of images or regions. However, such labels are typically too expensive to acquire, which restricts the scalability of supervised dictionary learning approaches. In this paper, we propose to leverage image attributes to weakly supervise the dictionary learning procedure without requiring any actual labels. As a key contribution, our approach establishes a generative hidden Markov random field (HMRF), which models the quantized codewords as the observed states and the image attributes as the hidden states, respectively. Dictionary learning is then performed by supervised grouping the observed states, where the supervised information is stemmed from the hidden states of the HMRF. In such a way, the proposed dictionary learning approach incorporates the image attributes to learn a semantic-preserving BoF representation without any genuine supervision. Experiments in large-scale image retrieval and classification tasks corroborate that our approach significantly outperforms the state-of-the-art unsupervised dictionary learning approaches.

Visual reranking has been widely deployed to refine the quality of conventional content-based image retrieval engines. The current trend lies in employing a crowd of retrieved results stemming from multiple feature modalities to boost the overall performance of visual reranking. However, a major challenge pertaining to current reranking methods is how to take full advantage of the complementary property of distinct feature modalities. Given a query image and one feature modality, a regular visual reranking framework treats the top-ranked images as pseudo positive instances which are inevitably noisy, difficult to reveal this complementary property, and thus lead to inferior ranking performance. This paper proposes a novel image reranking approach by introducing a Co-Regularized Multi-Graph Learning (Co-RMGL) framework, in which the intra-graph and inter-graph constraints are simultaneously imposed to encode affinities in a single graph and consistency across different graphs. Moreover, weakly supervised learning driven by image attributes is performed to denoise the pseudo-labeled instances, thereby highlighting the unique strength of individual feature modality. Meanwhile, such learning can yield a few anchors in graphs that vitally enable the alignment and fusion of multiple graphs. As a result, an edge weight matrix learned from the fused graph automatically gives the ordering to the initially retrieved results. We evaluate our approach on four benchmark image retrieval datasets, demonstrating a significant performance gain over the state-of-the-arts.

Hashing techniques have been intensively investigated in the design of highly efficient search engines for large-scale computer vision applications. Compared with prior approximate nearest neighbor search approaches like tree-based indexing, hashing-based search schemes have prominent advantages in terms of both storage and computational efficiencies. Moreover, the procedure of devising hash functions can be easily incorporated into sophisticated machine learning tools, leading to data-dependent and task-specific compact hash codes. Therefore, a number of learning paradigms, ranging from unsupervised to supervised, have been applied to compose appropriate hash functions. However, most of the existing hash function learning methods either treat hash function design as a classification problem or generate binary codes to satisfy pair wise supervision, and have not yet directly optimized the search accuracy. In this paper, we propose to leverage list wise supervision into a principled hash function learning framework. In particular, the ranking information is represented by a set of rank triplets that can be used to assess the quality of ranking. Simple linear projection-based hash functions are solved efficiently through maximizing the ranking quality over the training data. We carry out experiments on large image datasets with size up to one million and compare with the state-of-the-art hashing techniques. The extensive results corroborate that our learned hash codes via list wise supervision can provide superior search accuracy without incurring heavy computational overhead.

The archetype hull model is playing an important role in large-scale data analytics and mining, but rarely applied to vision problems. In this paper, we migrate such a geometric model to address face recognition and verification together through proposing a unified archetype hull ranking framework. Upon a scalable graph characterized by a compact set of archetype exemplars whose convex hull encompasses most of the training images, the proposed framework explicitly captures the relevance between any query and the stored archetypes, yielding a rank vector over the archetype hull. The archetype hull ranking is then executed for every block of face images to generate a block wise similarity measure that is achieved by comparing two different rank vectors with respect to the same archetype hull. After integrating block wise similarity measurements with learned importance weights, we accomplish a sensible face similarity measure which can support robust and effective face recognition and verification. We evaluate the face similarity measure in terms of experiments performed on three benchmark face databases Multi-PIE, Pubfig83, and LFW, demonstrating its performance superior to the state-of-the-arts.

Although text-based image search engines are popular for ranking images of user's interest, the state-of-the-art ranking performance is still far from satisfactory. One major issue comes from the visual similarity metric used in the ranking operation, which depends solely on visual features. To tackle this issue, one feasible method is to incorporate semantic concepts, also known as image attributes, into image ranking. However, the optimal combination of visual features and image attributes remains unknown. In this paper, we propose a query-dependent image reranking approach by leveraging the higher level attribute detection among the top returned images to adapt the dictionary built over the visual features to a query-specific fashion. We start from offline learning transposition probabilities between visual codewords and attributes, then utilize the probabilities to online adapt the dictionary, and finally produce a query-dependent and semantics-induced metric for image ranking. Extensive evaluations on several benchmark image datasets demonstrate the effectiveness and efficiency of the proposed approach in comparison with state-of-the-arts.

Recently, nonnegative matrix factorization (NMF) has become increasingly popular for feature extraction in computer vision and pattern recognition. NMF seeks two nonnegative matrices whose product can best approximate the original matrix. The nonnegativity constraints lead to sparse parts-based representations that can be more robust than nonsparse global features. To obtain more accurate control over the sparseness, in this paper, we propose a novel method called nonnegative local coordinate factorization (NLCF) for feature extraction. NLCF adds a local coordinate constraint into the standard NMF objective function. Specifically, we require that the learned basis vectors be as close to the original data points as possible. In this way, each data point can be represented by a linear combination of only a few nearby basis vectors, which naturally leads to sparse representation. Extensive experimental results suggest that the proposed approach provides a better representation and achieves higher accuracy in image clustering.

Recent years have witnessed the growing popularity of hashing in large-scale vision problems. It has been shown that the hashing quality could be boosted by leveraging supervised information into hash function learning. However, the existing supervised methods either lack adequate performance or often incur cumbersome model training. In this paper, we propose a novel kernel-based supervised hashing model which requires a limited amount of supervised information, i.e., similar and dissimilar data pairs, and a feasible training cost in achieving high quality hashing. The idea is to map the data to compact binary codes whose Hamming distances are minimized on similar pairs and simultaneously maximized on dissimilar pairs. Our approach is distinct from prior works by utilizing the equivalence between optimizing the code inner products and the Hamming distances. This enables us to sequentially and efficiently train the hash functions one bit at a time, yielding very short yet discriminative codes. We carry out extensive experiments on two image benchmarks with up to one million samples, demonstrating that our approach significantly outperforms the state-of-the-arts in searching both metric distance neighbors and semantically similar neighbors, with accuracy gains ranging from 13% to 46%.

Recent years have witnessed an explosive growth of user contributed videos on websites like YouTube and Metacafe, which usually provide a query-by-keyword functionality to facilitate the user browsing. For a given query, the returned videos typically contain multiple topics that are mixed up to duplicate the user browsing. Therefore, their diversification and grouping are highly demanded to improve the user experiences. However, the tagging and content qualities of user contributed videos are uncontrolled against their precise grouping. In this paper, we present a weakly supervised topic grouping paradigm to diversify the returned videos of a given keyword query. Our grouping is based on the bag-of-words visual signature quantized over the spatiotemporal STIP descriptor [1] extracted from each returned video. First, we adopt a min-Hashing based visual similarity in combination of the tagging similarity to group the returned videos. Based on the initial grouping configurations, we mine the co-occurred discriminative sub-signatures, based on which we iteratively refine the first step. Such iteration well handles the noise in visual content and tagging, since neither of which is fully trusted during the grouping. We validate our schemes on over 2,000 video clips crawled from a set of YouTube keyword query results. Comparing to alternative approaches, our scheme has shown superior robustness and precision.

In a typical content-based image retrieval (CBIR) system, images are represented as vectors and similarities between images are measured by a specified distance metric. However, the traditional Euclidean distance cannot always deliver satisfactory performance, so an effective metric sensible to the input data is desired. Tremendous recent works on metric learning have exhibited promising performance, but most of them suffer from limited label information and expensive training costs. In this paper, we propose two novel metric learning approaches, Optimal Semi-Supervised Metric Learning and its kernelized version. In the proposed approaches, we incorporate information from both labeled and unlabeled data to design a convex and computationally tractable learning framework which results in a globally optimal solution to the target metric of much lower rank than the original data dimension. Experiments on several image benchmarks demonstrate that our approaches lead to consistently better distance metrics than the state-of-the-arts in terms of accuracy for image retrieval.

Graph-based semisupervised learning (GSSL) provides a promising paradigm for modeling the manifold structures that may exist in massive data sources in high-dimensional spaces. It has been shown effective in propagating a limited amount of initial labels to a large amount of unlabeled data, matching the needs of many emerging applications such as image annotation and information retrieval. In this paper, we provide reviews of several classical GSSL methods and a few promising methods in handling challenging issues often encountered in web-scale applications. First, to successfully incorporate the contaminated noisy labels associated with web data, label diagnosis and tuning techniques applied to GSSL are surveyed. Second, to support scalability to the gigantic scale (millions or billions of samples), recent solutions based on anchor graphs are reviewed. To help researchers pursue new ideas in this area, we also summarize a few popular data sets and software tools publicly available. Important open issues are discussed at the end to stimulate future research.

A visual codebook serves as a fundamental component in many state-of-the-art computer vision systems. Most existing codebooks are built based on quantizing local feature descriptors extracted from training images. Subsequently, each image is represented as a high-dimensional bag-of-words histogram. Such highly redundant image description lacks efficiency in both storage and retrieval, in which only a few bins are nonzero and distributed sparsely. Furthermore, most existing codebooks are built based solely on the visual statistics of local descriptors, without considering the supervise labels coming from the subsequent recognition or classification tasks. In this paper, we propose a task-dependent codebook compression framework to handle the above two problems. First, we propose to learn a compression function to map an originally high-dimensional codebook into a compact codebook while maintaining its visual discriminability. This is achieved by a codeword sparse coding scheme with Lasso regression, which minimizes the descriptor distortions of training images after codebook compression. Second, we propose to adapt our codebook compression to the subsequent recognition or classification tasks. This is achieved by introducing a label constraint kernel (LCK) into our compression loss function. In particular, our LCK can model heterogeneous kinds of supervision, i.e., (partial) category labels, correlative semantic annotations, and image query logs. We validated our codebook compression in three computer vision tasks: 1) object recognition in PASCAL Visual Object Class 07; 2) near-duplicate image retrieval in UKBench; and 3) web image search in a collection of 0.5 million Flickr photographs. Our compressed codebook has shown superior performances over several state-of-the-art supervised and unsupervised codebooks.

In this paper, we exploit a novel ranking mechanism that processes query samples with noisy labels, motivated by the practical application of web image search re-ranking where the originally highest ranked images are usually posed as pseudo queries for subsequent re-ranking. Availing ourselves of the low-frequency spectrum of a neighborhood graph built on the samples, we propose a graph-theoretical framework amenable to noise resistant ranking. The proposed framework consists of two components: spectral filtering and graph-based ranking. The former leverages sparse bases, progressively selected from a pool of smooth eigenvectors of the graph Laplacian, to reconstruct the noisy label vector associated with the query sample set and accordingly filter out the query samples with less authentic positive labels. The latter applies a canonical graph ranking algorithm with respect to the filtered query sample set. Quantitative image re-ranking experiments carried out on two public web image databases bear out that our re-ranking approach compares favorably with the state-of-the-arts and improves web image search engines by a large margin though we harvest the noisy queries from the top-ranked images returned by these search engines.

Manifold Ranking (MR), a graph-based ranking algorithm, has been widely applied in information retrieval and shown to have excellent performance and feasibility on a variety of data types. Particularly, it has been successfully applied to content-based image retrieval, because of its outstanding ability to discover underlying geometrical structure of the given image database. However, manifold ranking is computationally very expensive, both in graph construction and ranking computation stages, which significantly limits its applicability to very large data sets. In this paper, we extend the original manifold ranking algorithm and propose a new framework named Efficient Manifold Ranking (EMR). We aim to address the shortcomings of MR from two perspectives: scalable graph construction and efficient computation. Specifically, we build an anchor graph on the data set instead of the traditional k-nearest neighbor graph, and design a new form of adjacency matrix utilized to speed up the ranking computation. The experimental results on a real world image database demonstrate the effectiveness and efficiency of our proposed method. With a comparable performance to the original manifold ranking, our method significantly reduces the computational time, makes it a promising method to large scale real world retrieval problems.

Distribution calibration plays an important role in cross-domain learning. However, existing distribution distance metrics are not geodesic; therefore, they cannot measure the intrinsic distance between two distributions. In this paper, we calibrate two distributions by using the geodesic distance in Riemannian symmetric space. Our method learns a latent subspace in the reproducing kernel Hilbert space, where the geodesic distance between the distribution of the source and the target domains is minimized. The corresponding geodesic distance is thus equivalent to the geodesic distance between two symmetric positive definite (SPD) matrices defined in the Riemannian symmetric space. These two SPD matrices parameterize the marginal distributions of the source and target domains in the latent subspace. We carefully design an evolutionary algorithm to find a local optimal solution that minimizes this geodesic distance. Empirical studies on face recognition, text categorization, and web image annotation suggest the effectiveness of the proposed scheme.

Vectored data frequently occur in a variety of fields, which are easy to handle since they can be mathematically abstracted as points residing in a Euclidean space. An appropriate distance metric in the data space is quite demanding for a great number of applications. In this paper, we pose robust and tractable metric learning under pairwise constraints that are expressed as similarity judgements between data pairs. The major features of our approach include: 1) it maximizes the gap between the average squared distance among dissimilar pairs and the average squared distance among similar pairs; 2) it is capable of propagating similar constraints to all data pairs; and 3) it is easy to implement in contrast to the existing approaches using expensive optimization such as semidefinite programming. Our constrained metric learning approach has widespread applicability without being limited to particular backgrounds. Quantitative experiments are performed for classification and retrieval tasks, uncovering the effectiveness of the proposed approach.

Scalable similarity search is the core of many large scale learning or data mining applications. Recently, many research results demonstrate that one promising approach is creating compact and efficient hash codes that preserve data similarity. By efficient, we refer to the low correlation (and thus low redundancy) among generated codes. However, most existing hash methods are designed only for vector data. In this paper, we develop a new hashing algorithm to create efficient codes for large scale data of general formats with any kernel function, including kernels on vectors, graphs, sequences, sets and so on. Starting with the idea analogous to spectral hashing, novel formulations and solutions are proposed such that a kernel based hash function can be explicitly represented and optimized, and directly applied to compute compact hash codes for new samples of general formats. Moreover, we incorporate efficient techniques, such as Nystrom approximation, to further reduce time and space complexity for indexing and search, making our algorithm scalable to huge data sets. Another important advantage of our method is the ability to handle diverse types of similarities according to actual task requirements, including both feature similarities and semantic similarities like label consistency. We evaluate our method using both vector and non-vector data sets at a large scale up to 1 million samples. Our comprehensive results show the proposed method outperforms several state-of-the-art approaches for all the tasks, with a significant gain for most tasks.

In plenty of scenarios, data can be represented as vectors and then mathematically abstracted as points in a Euclidean space. Because a great number of machine learning and data mining applications need proximity measures over data, a simple and universal distance metric is desirable, and metric learning methods have been explored to produce sensible distance measures consistent with data relationship. However, most existing methods suffer from limited labeled data and expensive training. In this paper, we address these two issues through employing abundant unlabeled data and pursuing sparsity of metrics, resulting in a novel metric learning approach called semi-supervised sparse metric learning. Two important contributions of our approach are: 1) it propagates scarce prior affinities between data to the global scope and incorporates the full affinities into the metric learning; and 2) it uses an efficient alternating linearization method to directly optimize the sparse metric. Compared with conventional methods, ours can effectively take advantage of semi-supervision and automatically discover the sparse metric structure underlying input data patterns. We demonstrate the efficacy of the proposed approach with extensive experiments carried out on six datasets, obtaining clear performance gains over the state-of-the-arts.

Learning a good distance metric plays a vital role in many multimedia retrieval and data mining tasks. For example, a typical content-based image retrieval (CBIR) system often relies on an effective distance metric to measure similarity between any two images. Conventional CBIR systems simply adopting Euclidean distance metric often fail to return satisfactory results mainly due to the well-known semantic gap challenge. In this article, we present a novel framework of Semi-Supervised Distance Metric Learning for learning effective distance metrics by exploring the historical relevance feedback log data of a CBIR system and utilizing unlabeled data when log data are limited and noisy. We formally formulate the learning problem into a convex optimization task and then present a new technique, named as “Laplacian Regularized Metric Learning” (LRML). Two efficient algorithms are then proposed to solve the LRML task. Further, we apply the proposed technique to two applications. One direct application is for Collaborative Image Retrieval (CIR), which aims to explore the CBIR log data for improving the retrieval performance of CBIR systems. The other application is for Collaborative Image Clustering (CIC), which aims to explore the CBIR log data for enhancing the clustering performance of image pattern clustering tasks. We conduct extensive evaluation to compare the proposed LRML method with a number of competing methods, including 2 standard metrics, 3 unsupervised metrics, and 4 supervised metrics with side information. Encouraging results validate the effectiveness of the proposed technique.

Graph-based methods form a main category of semi-supervised learning, offering flexibility and easy implementation in many applications. However, the performance of these methods is often sensitive to the construction of a neighborhood graph, which is non-trivial for many real-world problems. In this paper, we propose a novel framework that builds on learning the graph given labeled and unlabeled data. The paper has two major contributions. Firstly, we use a nonparametric algorithm to learn the entire adjacency matrix of a symmetry-favored k-NN graph, assuming that the matrix is doubly stochastic. The nonparametric algorithm makes the constructed graph highly robust to noisy samples and capable of approximating underlying submanifolds or clusters. Secondly, to address multi-class semi-supervised classification, we formulate a constrained label propagation problem on the learned graph by incorporating class priors, leading to a simple closed-form solution. Experimental results on both synthetic and real-world datasets show that our approach is significantly better than the state-of-the-art graph-based semi-supervised learning algorithms in terms of accuracy and robustness.

Image search is becoming prevalent in web search as the number of digital photos grows exponentially on the internet. For a successful image search system, removing outliers in the top ranked results is a challenging task. Typical content based image search engines take an input image from one class as a query and compute relevance between the query and images in a database. The results often contain a large number of outliers, since these outliers may be similar to the query image in some way. In this paper we present a novel search scheme using query images from multiple classes. Instead of conducting query search for one image class at a time, we conduct multi-class query search jointly. By using several query classes that are similar to each other for multi-class query, we can utilize information across similar classes to fine tune the similarity measure to remove outliers. This strategy can be used for any information search application. In this work, we use content based image search to illustrate the concept.

Typical content-based image retrieval (CBIR) solutions with regular Euclidean metric usually cannot achieve satisfactory performance due to the semantic gap challenge. Hence, relevance feedback has been adopted as a promising approach to improve the search performance. In this paper, we propose a novel idea of learning with historical relevance feedback log data, and adopt a new paradigm called ldquoCollaborative Image Retrievalrdquo (CIR). To effectively explore the log data, we propose a novel semi-supervised distance metric learning technique, called ldquoLaplacian Regularized Metric Learningrdquo (LRML), for learning robust distance metrics for CIR. Different from previous methods, the proposed LRML method integrates both log data and unlabeled data information through an effective graph regularization framework. We show that reliable metrics can be learned from real log data even they may be noisy and limited at the beginning stage of a CIR system. We conducted extensive evaluation to compare the proposed method with a large number of competing methods, including 2 standard metrics, 3 unsupervised metrics, and 4 supervised metrics with side information.

In this paper, we study semisupervised linear dimensionality reduction. Beyond conventional supervised methods which merely consider labeled instances, the semisupervised scheme allows to leverage abundant and ample unlabeled instances into learning so as to achieve better generalization performance. Under semisupervised settings, our objective is to learn a smooth as well as discriminative subspace and linear dimensionality reduction is thus achieved by mapping all samples into the subspace. Specifically, we present the transductive component analysis (TCA) algorithm to generate such a subspace founded on a graph-theoretic framework. Considering TCA is nonorthogonal, we further present the orthogonal transductive component analysis (OTCA) algorithm to iteratively produce a series of orthogonal basis vectors. OTCA has better discriminating power than TCA. Experiments carried out on synthetic and real-world datasets by OTCA show a clear improvement over the results of representative dimensionality reduction algorithms.

Relevant Component Analysis (RCA) has been proposed for learning distance metrics with contextual constraints for image retrieval. However, RCA has two important disadvantages. One is the lack of exploiting negative constraints which can also be informative, and the other is its incapability of capturing complex nonlinear relationships between data instances with the contextual information. In this paper, we propose two algorithms to overcome these two disadvantages, i.e., Discriminative Component Analysis (DCA) and Kernel DCA. Compared with other complicated methods for distance metric learning, our algorithms are rather simple to understand and very easy to solve. We evaluate the performance of our algorithms on image retrieval in which experimental results show that our algorithms are effective and promising in learning good quality distance metrics for image retrieval.