Multimodal Retrieval-Augmented Generation (MRAG) has recently been explored to empower Large Vision Language Models (LVLMs) with more comprehensive and up-to-date contextual knowledge, aiming to compensate for their limited and coarse-grained parametric knowledge in knowledge-intensive tasks. However, the retrieved contextual knowledge is usually not aligned with LVLMs’ internal parametric knowledge, leading to knowledge conflicts and further unreliable responses. To tackle this issue, we design KCM, a training-free and plug-and-play framework that can effectively mitigate knowledge conflicts while incorporating MRAG for more accurate LVLM responses. KCM enhances contextual knowledge utilization by modifying the LVLM architecture from three key perspectives. First, KCM adaptively adjusts attention distributions among multiple attention heads, encouraging LVLMs to focus on contextual knowledge with reduced distraction. Second, KCM identifies and prunes knowledge-centric LVLM neurons that encode coarse-grained parametric knowledge, thereby suppressing interferences and enabling more effective integration of contextual knowledge. Third, KCM amplifies the information flow from the input context by injecting supplementary context logits, reinforcing its contribution to the final output. Extensive experiments over multiple LVLMs and benchmarks show that KCM outperforms the state-of-the-art consistently by large margins, incurring neither extra training nor external tools.

Sparse-view 3D Gaussian splatting seeks to render high-quality novel views of 3D scenes from a limited set of input images. While recent pose-free feed-forward methods leveraging pre-trained 3D priors have achieved impressive results, most of them rely on full fine-tuning of large Vision Transformer (ViT) backbones and incur substantial GPU costs. In this work, we introduce MuSASplat, a novel framework that dramatically reduces the computational burden of training pose-free feed-forward 3D Gaussian splats models with little compromise of rendering quality. Central to our approach is a lightweight Multi-Scale Adapter that enables efficient fine-tuning of ViT-based architectures with only a small fraction of training parameters. This design avoids the prohibitive GPU overhead associated with previous full-model adaptation techniques while maintaining high fidelity in novel view synthesis, even with very sparse input views. In addition, we introduce a Feature Fusion Aggregator that integrates features across input views effectively and efficiently. Unlike widely adopted memory banks, the Feature Fusion Aggregator ensures consistent geometric integration across input views and meanwhile mitigates the memory usage, training complexity, and computational costs significantly. Extensive experiments across diverse datasets show that MuSASplat achieves state-of-the-art rendering quality but has significantly reduced parameters and training resource requirements as compared with existing methods.

Transformers have been successfully applied in the field of video-based 3D human pose estimation. However, the high computational costs of these video pose transformers (VPTs) make them impractical on resource-constrained devices. In this paper, we present a hierarchical plug-and-play pruning-and-recovering framework, calledHierarchicalHourglassTokenizer (H2OT), for efficient transformer-based 3D human pose estimation from videos. H2OT begins with progressively pruning pose tokens of redundant frames and ends with recovering full-length sequences, resulting in a few pose tokens in the intermediate transformer blocks and thus improving the model efficiency. It works with two key modules, namely, a Token Pruning Module (TPM) and a Token Recovering Module (TRM). TPM dynamically selects a few representative tokens to eliminate the redundancy of video frames, while TRM restores the detailed spatio-temporal information based on the selected tokens, thereby expanding the network output to the original full-length temporal resolution for fast inference. Our method is general-purpose: it can be easily incorporated into common VPT models on bothseq2seqandseq2framepipelines while effectively accommodating different token pruning and recovery strategies. In addition, our H2OT reveals that maintaining the full pose sequence is unnecessary, and a few pose tokens of representative frames can achieve both high efficiency and estimation accuracy. Extensive experiments on multiple benchmark datasets demonstrate both the effectiveness and efficiency of the proposed method. Code and models are available athttps://github.com/NationalGAILab/HoT.

No-Reference Image Quality Assessment (NR-IQA), responsible for assessing the quality of a single input image without using any reference, plays a critical role in evaluating and optimizing computer vision systems, e.g., low-light enhancement. Recent research indicates that NR-IQA models are susceptible to adversarial attacks, which can significantly alter predicted scores with visually imperceptible perturbations. Despite revealing vulnerabilities, these attack methods have limitations, including high computational demands, untargeted manipulation, limited practical utility in white-box scenarios, and reduced effectiveness in black-box scenarios. To address these challenges, we shift our focus to another significant threat and present a novel poisoning-based backdoor attack against NR-IQA (BAIQA), allowing the attacker to manipulate the IQA model's output to any desired target value by simply adjusting a scaling coefficient alpha for the trigger. We propose to inject the trigger in the discrete cosine transform (DCT) domain to improve the local invariance of the trigger for countering trigger diminishment in NR-IQA models due to widely adopted data augmentations. Furthermore, the universal adversarial perturbations (UAP) in the DCT space are designed as the trigger, to increase IQA model susceptibility to manipulation and improve attack effectiveness. In addition to the heuristic method for poison-label BAIQA (P-BAIQA), we explore the design of clean-label BAIQA (C-BAIQA), focusing on alpha sampling and image data refinement, driven by theoretical insights we reveal. Extensive experiments on diverse datasets and various NR-IQA models demonstrate the effectiveness of our attacks.

Despite great success across various multimodal tasks, Large Vision-Language Models (LVLMs) often encounter object hallucinations with generated textual responses being inconsistent with the actual objects in images. We examine different LVLMs and pinpoint that one root cause of object hallucinations lies with deficient attention on discriminative image features. Specifically, LVLMs often predominantly attend to prompt-irrelevant global features instead of prompt-relevant local features, undermining their visual grounding capacity and leading to object hallucinations. We propose Assembly of Global and Local Attention (AGLA), a training-free and plug-and-play approach that mitigates hallucinations by assembling global features for response generation and local features for visual discrimination simultaneously. Specifically, we introduce an image-prompt matching scheme that captures prompt-relevant local features from images, leading to an augmented view of the input image where prompt-relevant content is highlighted while irrelevant distractions are suppressed. Hallucinations can thus be mitigated with a calibrated logit distribution that is from generative global features of the original image and discriminative local features of the augmented image. Extensive experiments show the superiority of AGLA in LVLM hallucination mitigation, demonstrating its wide applicability across both discriminative and generative tasks. Our code is available at https://github.com/Lackel/AGLA.

Anchor-based 3D Gaussian splatting (3D-GS) exploits anchor features in 3D Gaussian prediction, which has achieved impressive 3D rendering quality with reduced Gaussian redundancy. On the other hand, it often encounters the dilemma among anchor features, model size, and rendering quality – large anchor features lead to large 3D models and high-quality rendering whereas reducing anchor features degrades Gaussian attribute prediction which leads to clear artifacts in the rendered textures and geometries. We design SOGS, an anchor-based 3D-GS technique that introduces second-order anchors to achieve superior rendering quality and reduced anchor features and model size simultaneously. Specifically, SOGS incorporates covariance-based second-order statistics and correlation across feature dimensions to augment features within each anchor, compensating for the reduced feature size and improving rendering quality effectively. In addition, it introduces a selective gradient loss to enhance the optimization of scene textures and scene geometries, leading to high-quality rendering with small anchor features. Extensive experiments over multiple widely adopted benchmarks show that SOGS achieves superior rendering quality in novel view synthesis with clearly reduced model size.

Weihao Xuan, Rui Yang, Heli Qi, Qingcheng Zeng, Yunze Xiao, Aosong Feng, Dairui Liu, Yun Xing, Junjue Wang, Fan Gao, Jinghui Lu, Yuang Jiang, Huitao Li, Xin Li, Kunyu Yu, Ruihai Dong, Shangding Gu, Yuekang Li, Xiaofei Xie, Felix Juefei-Xu, Foutse Khomh, Osamu Yoshie, Qingyu Chen, Douglas Teodoro, Nan Liu, Randy Goebel, Lei Ma, Edison Marrese-Taylor, Shijian Lu, Yusuke Iwasawa, Yutaka Matsuo, Irene Li. Proceedings of the 2025 Conference on Empirical Methods in Natural Language Processing. 2025.

Generalizable depth completion enables the acquisition of dense metric depth maps for unseen environments, offering robust perception capabilities for various downstream tasks. However, training such models typically requires large-scale datasets with metric depth labels, which are often labor-intensive to collect. This paper presents PacGDC, a label-efficient technique that enhances data diversity with minimal annotation effort for generalizable depth completion. PacGDC builds on novel insights into inherent ambiguities and consistencies in object shapes and positions during 2D-to-3D projection, allowing the synthesis of numerous pseudo geometries for the same visual scene. This process greatly broadens available geometries by manipulating scene scales of the corresponding depth maps. To leverage this property, we propose a new data synthesis pipeline that uses multiple depth foundation models as scale manipulators. These models robustly provide pseudo depth labels with varied scene scales, affecting both local objects and global layouts, while ensuring projection consistency that supports generalization. To further diversify geometries, we incorporate interpolation and relocation strategies, as well as unlabeled images, extending the data coverage beyond the individual use of foundation models. Extensive experiments show that PacGDC achieves remarkable generalizability across multiple benchmarks, excelling in diverse scene semantics/scales and depth sparsity/patterns under both zero-shot and few-shot settings. Code: https://github.com/Wang-xjtu/PacGDC.

COLMAP-free 3D Gaussian Splatting (3D-GS) has recently attracted increasing attention due to its remarkable performance in reconstructing high-quality 3D scenes from unposed images or videos. However, it often struggles to handle scenes with complex camera trajectories as featured by drastic rotation and translation across adjacent camera views, leading to degraded estimation of camera poses and further local minima in joint optimization of camera poses and 3D-GS. We propose PCR-GS, an innovative COLMAP-free 3DGS technique that achieves superior 3D scene modeling and camera pose estimation via camera pose co-regularization. PCR-GS achieves regularization from two perspectives. The first is feature reprojection regularization which extracts view-robust DINO features from adjacent camera views and aligns their semantic information for camera pose regularization. The second is wavelet-based frequency regularization which exploits discrepancy in high-frequency details to further optimize the rotation matrix in camera poses. Extensive experiments over multiple real-world scenes show that the proposed PCR-GS achieves superior pose-free 3D-GS scene modeling under dramatic changes of camera trajectories.

Video Temporal Grounding (VTG) aims to precisely identify video event segments in response to textual queries. The outputs of VTG tasks manifest as sequences of events, each defined by precise timestamps, saliency scores, and textual descriptions. Despite recent advances, a fundamental limitation persists in existing Video Large Language Models (Video-LLMs): they process all task tokens through identical and static pathways, failing to recognize that temporal localization, saliency assessment, and textual generation represent fundamentally distinct tasks requiring specialized processing. To address this, we introduce TimeExpert, a Mixture-of-Experts (MoE)-based Video-LLM that effectively decomposes VTG tasks by dynamically routing task-specific tokens (e.g., timestamps, saliency scores) to specialized experts, with increased computational efficiency. Our design choices enable precise handling of each subtask, leading to improved event modeling across diverse VTG applications. Extensive experiments demonstrate that TimeExpert consistently achieves state-of-the-art performance on various VTG tasks such as Dense Video Captioning, Moment Retrieval, and Video Highlight Detection.

Leveraging text, images, structure maps, or motion trajectories as conditional guidance, diffusion models have achieved great success in automated and high-quality video generation. However, generating smooth and rational transition videos given the first and last video frames as well as descriptive text prompts is far underexplored. We present VTG, a Versatile Transition video Generation framework that can generate smooth, high-fidelity, and semantically coherent video transitions. VTG introduces interpolation-based initialization that helps preserve object identity and handle abrupt content changes effectively. In addition, it incorporates dual-directional motion fine-tuning and representation alignment regularization to mitigate the limitations of pre-trained image-to-video diffusion models in motion smoothness and generation fidelity, respectively. To evaluate VTG and facilitate future studies on unified transition generation, we collected TransitBench, a comprehensive benchmark for transition generation covering two representative transition tasks: concept blending and scene transition. Extensive experiments show that VTG achieves superior transition performance consistently across all four tasks.

Recent studies generally enhance MLLMs' reasoning capabilities via supervised fine-tuning on high-quality chain-of-thought reasoning data, which often leads models to merely imitate successful reasoning paths without understanding what the wrong reasoning paths are. In this work, we aim to enhance the MLLMs' reasoning ability beyond passively imitating positive reasoning paths. To this end, we design Step-wise Group Relative Policy Optimization (StepGRPO), a new online reinforcement learning framework that enables MLLMs to self-improve reasoning ability via simple, effective and dense step-wise rewarding. Specifically, StepGRPO introduces two novel rule-based reasoning rewards: Step-wise Reasoning Accuracy Reward (StepRAR) and Step-wise Reasoning Validity Reward (StepRVR). StepRAR rewards the reasoning paths that contain necessary intermediate reasoning steps via a soft key-step matching technique, while StepRAR rewards reasoning paths that follow a well-structured and logically consistent reasoning process through a reasoning completeness and logic evaluation strategy. With the proposed StepGRPO, we introduce R1-VL, a series of MLLMs with outstanding capabilities in step-by-step reasoning. Extensive experiments over 8 benchmarks demonstrate the superiority of our methods.

Most existing unlearnable strategies focus on preventing unauthorized users from training single-task learning (STL) models with personal data. Nevertheless, the paradigm has recently shifted towards multi-task data and multi-task learning (MTL), targeting generalist and foundation models that can handle multiple tasks simultaneously. Despite their growing importance, MTL data and models have been largely neglected while pursuing unlearnable strategies. This paper presents MTL-UE, the first unified framework for generating unlearnable examples for multi-task data and MTL models. Instead of optimizing perturbations for each sample, we design a generator-based structure that introduces label priors and class-wise feature embeddings which leads to much better attacking performance. In addition, MTL-UE incorporates intra-task and inter-task embedding regularization to increase inter-class separation and suppress intra-class variance which enhances the attack robustness greatly. Furthermore, MTL-UE is versatile with good supports for dense prediction tasks in MTL. It is also plug-and-play allowing integrating existing surrogate-dependent unlearnable methods with little adaptation. Extensive experiments show that MTL-UE achieves superior attacking performance consistently across 4 MTL datasets, 3 base UE methods, 5 model backbones, and 5 MTL task-weighting strategies. Code is available at https://github.com/yuyi-sd/MTL-UE.

Despite their recent progress, Multimodal Large Language Models (MLLMs) often struggle in knowledge-intensive tasks due to the limited and outdated parametric knowledge acquired during training. Multimodal Retrieval Augmented Generation addresses this issue by retrieving contextual knowledge from external databases, thereby enhancing MLLMs with expanded knowledge sources. However, existing MLLMs often fail to fully leverage the retrieved contextual knowledge for response generation. We examine representative MLLMs and identify two major causes, namely, attention bias toward different tokens and knowledge conflicts between parametric and contextual knowledge. To this end, we design Adaptive Logits Fusion and Attention Reallocation (ALFAR), a training-free and plug-and-play approach that improves MLLM responses by maximizing the utility of the retrieved knowledge. Specifically, ALFAR tackles the challenges from two perspectives. First, it alleviates attention bias by adaptively shifting attention from visual tokens to relevant context tokens according to query-context relevance. Second, it decouples and weights parametric and contextual knowledge at output logits, mitigating conflicts between the two types of knowledge. As a plug-and-play method, ALFAR achieves superior performance across diverse datasets without requiring additional training or external tools. Extensive experiments over multiple MLLMs and benchmarks show that ALFAR consistently outperforms the state-of-the-art by large margins. Our code and data are available at https://github.com/Lackel/ALFAR.

Recent advancements in large multimodal models (LMMs) have significantly enhanced performance across diverse tasks, with ongoing efforts to further integrate additional modalities such as video and audio. However, most existing LMMs remain vulnerable to hallucinations, the discrepancy between the factual multimodal input and the generated textual output, which has limited their applicability in various real-world scenarios. This paper presents the first systematic investigation of hallucinations in LMMs involving the three most common modalities: language, visual, and audio. Our study reveals two key contributors to hallucinations: overreliance on unimodal priors and spurious inter-modality correlations. To address these challenges, we introduce the benchmark The Curse of Multi-Modalities (CMM), which comprehensively evaluates hallucinations in LMMs, providing a detailed analysis of their underlying issues. Our findings highlight key vulnerabilities, including imbalances in modality integration and biases from training data, underscoring the need for balanced cross-modal learning and enhanced hallucination mitigation strategies. Based on our observations and findings, we suggest potential research directions that could enhance the reliability of LMMs.

As an essential vision task, infrared small target detection (IRSTD) has seen significant advancements through deep learning. However, critical limitations in current evaluation protocols impede further progress. First, existing methods rely on fragmented pixel- and target-level specific metrics, which fails to provide a comprehensive view of model capabilities. Second, an excessive emphasis on overall performance scores obscures crucial error analysis, which is vital for identifying failure modes and improving real-world system performance. Third, the field predominantly adopts dataset-specific training-testing paradigms, hindering the understanding of model robustness and generalization across diverse infrared scenarios. This paper addresses these issues by introducing a hybrid-level metric incorporating pixel- and target-level performance, proposing a systematic error analysis method, and emphasizing the importance of cross-dataset evaluation. These aim to offer a more thorough and rational hierarchical analysis framework, ultimately fostering the development of more effective and robust IRSTD models. An open-source toolkit has be released to facilitate standardized benchmarking.

Multi-modal image segmentation faces real-world deployment challenges from incomplete/corrupted modalities degrading performance. While existing methods address training-inference modality gaps via specialized per-combination models, they introduce high deployment costs by requiring exhaustive model subsets and model-modality matching. In this work, we propose a unified modality-relax segmentation network (UniMRSeg) through hierarchical self-supervised compensation (HSSC). Our approach hierarchically bridges representation gaps between complete and incomplete modalities across input, feature and output levels. First, we adopt modality reconstruction with the hybrid shuffled-masking augmentation, encouraging the model to learn the intrinsic modality characteristics and generate meaningful representations for missing modalities through cross-modal fusion. Next, modality-invariant contrastive learning implicitly compensates the feature space distance among incomplete-complete modality pairs. Furthermore, the proposed lightweight reverse attention adapter explicitly compensates for the weak perceptual semantics in the frozen encoder. Last, UniMRSeg is fine-tuned under the hybrid consistency constraint to ensure stable prediction under all modality combinations without large performance fluctuations. Without bells and whistles, UniMRSeg significantly outperforms the state-of-the-art methods under diverse missing modality scenarios on MRI-based brain tumor segmentation, RGB-D semantic segmentation, RGB-D/T salient object segmentation. The code will be released at \url{https://github.com/Xiaoqi-Zhao-DLUT/UniMRSeg}.

Text-to-Avatar generation has recently made significant strides due to advancements in diffusion models. However, most existing works remain constrained by limited diversity, producing avatars with subtle differences in appearance for a given text prompt. We design DivAvatar, a novel framework that generates diverse avatars, empowering 3D creatives with a multitude of distinct and richly varied 3D avatars from a single text prompt. Different from most existing works that exploit scene-specific 3D representations such as NeRF, DivAvatar finetunes a 3D generative model (i.e., EVA3D), allowing diverse avatar generation from simply noise sampling in inference time. DivAvatar has two key designs that help achieve generation diversity and visual quality. The first is a noise sampling technique during

Recent advancements in deep learning-based compression techniques have demonstrated remarkable performance surpassing traditional methods. Nevertheless, deep neural networks have been observed to be vulnerable to backdoor attacks, where an added pre-defined trigger pattern can induce the malicious behavior of the models. In this paper, we propose a novel approach to launch a backdoor attack with multiple triggers against learned image compression models. Drawing inspiration from the widely used discrete cosine transform (DCT) in existing compression codecs and standards, we propose a frequency-based trigger injection model that adds triggers in the DCT domain. In particular, we design several attack objectives that are adapted for a series of diverse scenarios, including: 1) attacking compression quality in terms of bit-rate and reconstruction quality; 2) attacking task-driven measures, such as face recognition and semantic segmentation in downstream applications. To facilitate more efficient training, we develop a dynamic loss function that dynamically balances the impact of different loss terms with fewer hyper-parameters, which also results in more effective optimization of the attack objectives with improved performance. Furthermore, we consider several advanced scenarios. We evaluate the resistance of the proposed backdoor attack to the defensive pre-processing methods and then propose a two-stage training schedule along with the design of robust frequency selection to further improve resistance. To strengthen both the cross-model and cross-domain transferability on attacking downstream CV tasks, we propose to shift the classification boundary in the attack loss during training. Extensive experiments also demonstrate that by employing our trained trigger injection models and making slight modifications to the encoder parameters of the compression model, our proposed attack can successfully inject multiple backdoors accompanied by their corresponding triggers into a single image compression model.

Few-shot object detection (FSOD) within remote sensing imagery has achieved great advancements in recent years. However, most existing methods are facing one key challenge while handling remote sensing images: many unlabeled instances in few-shot images are treated as background, which tends to degrade the generalization of the trained model severely. This paper presents HPMF, a Hypergraph-guided Prototype Mining Framework that addresses the challenge through joint optimization from three perspectives. The first is Hierarchical Reference Mining (HRM) which constructs a class-instance dual-driven prototype space that enables mining the unlabeled instances via cross-hierarchical similarity fusion. The second is a Robust Pseudo-box Estimator (RPE) that generates high-quality pseudo bounding boxes for the HRM-mined instances via adaptive density clustering and multi-statistic aggregation. The third is a Hypergraph-Guided Decoder (HGD) that introduces hypergraphs into the transformer decoder for group semantic modeling, enhancing high-order semantic association and similarity of instance features, thereby further improving the mining performance of the HRM module. Extensive experiments under various settings show that the proposed HPMF outperforms state-of-the-art methods consistently across multiple widely adopted remote-sensing FSOD benchmarks such as DIOR, NWPU-VHR10 v2, and HRRSD.

Recent advances in natural-domain multi-modal large language models (MLLMs) have demonstrated effective spatial reasoning through visual and textual prompting. However, their direct transfer to remote sensing (RS) is hindered by heterogeneous sensing physics, diverse modalities, and unique spatial scales. Existing RS MLLMs are mainly limited to optical imagery and plain language interaction, preventing flexible and scalable real-world applications. In this article, EarthGPT-X is proposed, the first flexible spatial MLLM that unifies multi-source RS imagery comprehension and accomplishes both coarse-grained and fine-grained visual tasks under diverse visual prompts in a single framework. Distinct from prior models, EarthGPT-X introduces: 1) a dual-prompt mechanism combining text instructions with various visual prompts (i.e., point, box, and free-form) to mimic the versatility of referring in human life; 2) a comprehensive multi-source multi-level prompting dataset, the model advances beyond holistic image understanding to support hierarchical spatial reasoning, including scene-level understanding and fine-grained object attributes and relational analysis; 3) a cross-domain one-stage fusion training strategy, enabling efficient and consistent alignment across modalities and tasks. Extensive experiments demonstrate that EarthGPT-X substantially outperforms prior natural and RS MLLMs, establishing the first framework capable of multi-source, multi-task, and multi-level interpretation using visual prompting in RS scenarios. The code and dataset are available athttps://github.com/wivizhang/EarthGPT-X.

Adversarial examples, characterized by imperceptible perturbations, pose significant threats to deep neural networks by misleading their predictions. A critical aspect of these examples is their transferability, allowing them to deceive unseen models in closed-box scenarios. Despite the widespread exploration of defense methods, including those on transferability, they show limitations: inefficient deployment, ineffective defense, and degraded performance on clean images. In this work, we introduce a novel training paradigm aimed at enhancing robustness against transferable adversarial examples (TAEs) in a more efficient and effective way. We propose a model that exhibits random guessing behavior when presented with clean data$\boldsymbol {x}$as input, and generates accurate predictions when with triggered data$\boldsymbol {x}+\boldsymbol {\tau }$. Importantly, the trigger$\boldsymbol {\tau }$remains constant for all data instances. We refer to these models as models with trigger activation. We are surprised to find that these models exhibit certain robustness against TAEs. Through the consideration of first-order gradients, we provide a theoretical analysis of this robustness. Moreover, through the joint optimization of the learnable trigger and the model, we achieve improved robustness to transferable attacks. Extensive experiments conducted across diverse datasets, evaluating a variety of attacking methods, underscore the effectiveness and superiority of our approach.

Multi-label image recognition with convolutional neural networks has achieved remarkable progress in the past few years. However, most existing multi-label image recognition methods suffer from the long-tailed data distribution problem,i.e., head categories occupy most training samples, while tailed classes have few samples. This work firstly studies the influence of long-tailed data distribution on existing multi-label image recognition methods. Based on this, two crucial issues of the existing methods are identified: 1) severe gradient imbalance between head and tailed categories, even though re-balancing strategies are adopted; 2) the lack of diversity of tail category training samples. To tackle the first issue, this paper proposes a group sampling strategy to create group-wise balanced data distribution. Meanwhile, a dynamic gradient balancing loss is proposed to equalize the gradient for all categories. To tackle the second issue, this paper proposes a diversity enhancement module to fuse the information across all categories, preventing the network from overfitting tail classes. Furthermore, it also balances the gradient, promoting the discriminability of learned classifiers. Our method significantly outperforms the baseline method and achieves competitive performance with state-of-the-art methods on VOC-LT and COCO-LT datasets. Extensive ablation studies are conducted to verify the effectiveness of the essential proposals.

Driven by rising demands in autonomous driving, robotics,etc., 3D object detection has recently achieved great advancement by fusing optical images and LiDAR point data. On the other hand, most existing optical-LiDAR fusion methods straightly overlay RGB images and point clouds without adequately exploiting the synergy between them, leading to suboptimal fusion and 3D detection performance. Additionally, they often suffer from limited localization accuracy without proper balancing of global and local object information. To address this issue, we design a synergistic network (SyNet) that fuses geometric information, semantic information, as well as global and local information of objects for robust and accurate 3D detection. The SyNet captures synergies between optical images and LiDAR point clouds from three perspectives. The first is geometric, which derives high-quality depth by projecting point clouds onto multi-view images, enriching optical RGB images with 3D spatial information for a more accurate interpretation of image semantics. The second is semantic, which voxelizes point clouds and establishes correspondences between the derived voxels and image pixels, enriching 3D point clouds with semantic information for more accurate 3D detection. The third is balancing local and global object information, which introduces deformable self-attention and cross-attention to process the two types of complementary information in parallel for more accurate object localization. Extensive experiments show that SyNet achieves 70.7% mAP and 73.5% NDS on the nuScenes test set, demonstrating its effectiveness and superiority as compared with the state-of-the-art.

The task of 3D single object tracking (SOT) with LiDAR point clouds is crucial for various applications, such as autonomous driving and robotics. However, existing approaches have primarily relied on appearance matching or motion modeling within only two successive frames, thereby overlooking the long-range continuous motion property of objects in 3D space. To address this issue, this paper presents a novel approach that views each tracklet as a continuous stream: at each timestamp, only the current frame is fed into the network to interact with multi-frame historical features stored in a memory bank, enabling efficient exploitation of sequential information. To achieve effective cross-frame message passing, a hybrid attention mechanism is designed to account for both long-range relation modeling and local geometric feature extraction. Furthermore, to enhance the utilization of multi-frame features for robust tracking, a contrastive sequence enhancement strategy is proposed, which uses ground truth tracklets to augment training sequences and promote discrimination against false positives in a contrastive manner. Extensive experiments demonstrate that the proposed method outperforms the state-of-the-art method by significant margins on multiple benchmarks.

Cross-Domain Few-Shot Segmentation (CD-FSS) poses the challenge of segmenting novel categories from a distinct domain using only limited exemplars. In this paper, we undertake a comprehensive study of CD-FSS and uncover two crucial insights: (i) the necessity of a fine-tuning stage to effectively transfer the learned meta-knowledge across domains, and (ii) the overfitting risk during the naive fine-tuning due to the scarcity of novel category examples. With these insights, we propose a novel cross-domain fine-tuning strategy that addresses the challenging CD-FSS tasks. We first design Bi-directional Few-shot Prediction (BFP), which establishes support-query correspondence in bi-directional manner, crafting augmented supervision to reduce the overfitting risk. Then we further extend BFP into Iterative Few-shot Adaptor (IFA), which is a recursive framework to capture the support-query correspondence iteratively, targeting maximal exploitation of supervisory signals from the sparse novel category samples. Extensive empirical evaluations show that our method significantly outperforms the state-of-the-arts (+7.8%), which verifies that IFA tackles the cross-domain challenges and mitigates the overfitting simultaneously.

Inspired by the success of general-purpose models in NLP, recent studies attempt to unify different vision tasks in the same sequence format and employ autoregressive Transformers for sequence prediction. They apply uni-directional attention to capture sequential dependencies and generate task sequences recursively. However, such autoregressive Transformers may not fit vision tasks well, as vision task sequences usually lack the sequential dependencies typically observed in natural languages. In this work, we design Masked AutoDecoder (MAD), an effective multitask vision generalist. MAD consists of two core designs. First, we develop a parallel decoding framework that introduces bi-directional attention to capture contextual dependencies comprehensively and decode vision task sequences in parallel. Second, we design a masked sequence modeling approach that learns rich task contexts by masking and reconstructing task sequences. In this way, MAD handles all the tasks by a single network branch and a simple cross-entropy loss with minimal task-specific designs. Extensive experiments demonstrate the great potential of MAD as a new paradigm for unifying various vision tasks. MAD achieves superior performance and inference efficiency compared to autoregressive counterparts while obtaining competitive accuracy with task-specific models. Code will be released at https://github.com/hanqiu-hq/MAD.

Monocular 3D detection (M3D) aims for precise 3D object localization from a single-view image which usually involves labor-intensive annotation of 3D detection boxes. Weakly supervised M3D has recently been studied to obviate the 3D annotation process by leveraging many existing 2D annotations, but it often requires extra training data such as LiDAR point clouds or multi-view images which greatly degrades its applicability and usability in various applications. We propose SKD-WM3D, a weakly supervised monocular 3D detection framework that exploits depth information to achieve M3D with a single-view image exclusively without any 3D annotations or other training data. One key design in SKD-WM3D is a self-knowledge distillation framework, which transforms image features into 3D-like representations by fusing depth information and effectively mitigates the inherent depth ambiguity in monocular scenarios with little computational overhead in inference. In addition, we design an uncertainty-aware distillation loss and a gradient-targeted transfer modulation strategy which facilitate knowledge acquisition and knowledge transfer, respectively. Extensive experiments show that SKD-WM3D surpasses the state-of-the-art clearly and is even on par with many fully supervised methods.

Test-time adaptation with pre-trained vision-language models has attracted increasing attention for tackling distribution shifts during the test time. Though prior studies have achieved very promising performance, they in-volve intensive computation which is severely unaligned with test-time adaptation. We design TDA, a training-free dynamic adapter that enables effective and efficient test-time adaptation with vision-language models. TDA works with a lightweight key-value cache that maintains a dy-namic queue with few-shot pseudo labels as values and the corresponding test-sample features as keys. Leveraging the key-value cache, TDA allows adapting to test data gradually via progressive pseudo label refinement which is super-efficient without incurring any backpropagation. In addition, we introduce negative pseudo labeling that alleviates the adverse impact of pseudo label noises by assigning pseudo labels to certain negative classes when the model is uncertain about its pseudo label predictions. Extensive experiments over two benchmarks demonstrate TDA's superior effectiveness and efficiency as compared with the state-of-the-art. The code has been released in https://kdiaaa.github.io/tda/

Large Vision-Language Models (LVLMs) have advanced considerably, intertwining visual recognition and language understanding to generate content that is not only coherent but also contextually attuned. Despite their success, LVLMs still suffer from the issue of object hallucinations, where models generate plausible yet incorrect outputs that include objects that do not exist in the images. To mitigate this issue, we introduce Visual Contrastive Decoding (VCD), a simple and training-free method that contrasts output distributions derived from original and distorted visual inputs. The proposed VCD effectively reduces the over-reliance on statistical bias and unimodal priors, two essential causes of object hallucinations. This adjustment ensures the generated content is closely grounded to visual inputs, resulting in contextually accurate outputs. Our experiments show that VCD, without either additional training or the usage of external tools, significantly mitigates the object hallucination issue across different LVLM families. Beyond mitigating object hallucinations, VCD also excels in general LVLM benchmarks, highlighting its wide-ranging applicability.

3D Gaussian splatting has achieved very impressive performance in real-time novel view synthesis. However, it often suffers from over-reconstruction during Gaussian densification where high-variance image regions are covered by a few large Gaussians only, leading to blur and artifacts in the rendered images. We design a progressive frequency regularization (FreGS) technique to tackle the over-reconstruction issue within the frequency space. Specifically, FreGS performs coarse-to-fine Gaussian densification by exploiting low-to-high frequency components that can be easily extracted with low-pass and high-pass filters in the Fourier space. By minimizing the discrepancy between the frequency spectrum of the rendered image and the corresponding ground truth, it achieves high-quality Gaussian densification and alleviates the over-reconstruction of Gaussian splatting effectively. Experiments over multiple widely adopted benchmarks (e.g., Mip-NeRF360, Tanks-and-Temples and Deep Blending) show that FreGS achieves superior novel view synthesis and outperforms the state-of-the-art consistently.

Inspired by the outstanding zero-shot capability of vision language models (VLMs) in image classification tasks, open-vocabulary object detection has attracted increasing interest by distilling the broad VLM knowledge into detector training. However, most existing open-vocabulary detectors learn by aligning region embeddings with categorical labels (e.g., bicycle) only, disregarding the capability of VLMs on aligning visual embeddings with fine-grained text descriptions of object parts (e.g., pedals and bells). This paper presents DVDet, a Descriptor-Enhanced Open Vocabulary Detector that introduces conditional context prompts and hierarchical textual descriptors that enable precise region-text alignment as well as open-vocabulary detection training in general. Specifically, the conditional context prompt transforms regional embeddings into image-like representations that can be directly integrated into general open vocabulary detection training. In addition, we introduce large language models as an interactive and implicit knowledge repository which enables iterative mining and refining visually oriented textual descriptors for precise region-text alignment. Extensive experiments over multiple large-scale benchmarks show that DVDet outperforms the state-of-the-art consistently by large margins.

Unlearnable examples (UEs) seek to maximize testing error by making subtle modifications to training examples that are correctly labeled. Defenses against these poisoning attacks can be categorized based on whether specific interventions are adopted during training. The first approach is training-time defense, such as adversarial training, which can mitigate poisoning effects but is computationally intensive. The other approach is pre-training purification, e.g., image short squeezing, which consists of several simple compressions but often encounters challenges in dealing with various UEs. Our work provides a novel disentanglement mechanism to build an efficient pre-training purification method. Firstly, we uncover rate-constrained variational autoencoders (VAEs), demonstrating a clear tendency to suppress the perturbations in UEs. We subsequently conduct a theoretical analysis for this phenomenon. Building upon these insights, we introduce a disentangle variational autoencoder (D-VAE), capable of disentangling the perturbations with learnable class-wise embeddings. Based on this network, a two-stage purification approach is naturally developed. The first stage focuses on roughly eliminating perturbations, while the second stage produces refined, poison-free results, ensuring effectiveness and robustness across various scenarios. Extensive experiments demonstrate the remarkable performance of our method across CIFAR-10, CIFAR-100, and a 100-class ImageNet-subset. Code is available at https://github.com/yuyi-sd/D-VAE.

Recent studies on generalizable object detection have attracted increasing attention with additional weak supervision from large-scale datasets with image-level labels. However, weakly-supervised detection learning often suffers from image-to-box label mismatch, i.e., image-level labels do not convey precise object information. We design Language Hierarchical Self-training (LHST) that introduces language hierarchy into weakly-supervised detector training for learning more generalizable detectors. LHST expands the image-level labels with language hierarchy and enables co-regularization between the expanded labels and self-training. Specifically, the expanded labels regularize self-training by providing richer supervision and mitigating the image-to-box label mismatch, while self-training allows assessing and selecting the expanded labels according to the predicted reliability. In addition, we design language hierarchical prompt generation that introduces language hierarchy into prompt generation which helps bridge the vocabulary gaps between training and testing. Extensive experiments show that the proposed techniques achieve superior generalization performance consistently across 14 widely studied object detection datasets.

Monocular 3D object detection aims for precise 3D localization and identification of objects from a single-view image. Despite its recent progress, it often struggles while handling pervasive object occlusions that tend to complicate and degrade the prediction of object dimensions, depths, and orientations. We design MonoMAE, a monocular 3D detector inspired by Masked Autoencoders that addresses the object occlusion issue by masking and reconstructing objects in the feature space. MonoMAE consists of two novel designs. The first is depth-aware masking that selectively masks certain parts of non-occluded object queries in the feature space for simulating occluded object queries for network training. It masks non-occluded object queries by balancing the masked and preserved query portions adaptively according to the depth information. The second is lightweight query completion that works with the depth-aware masking to learn to reconstruct and complete the masked object queries. With the proposed feature-space occlusion and completion, MonoMAE learns enriched 3D representations that achieve superior monocular 3D detection performance qualitatively and quantitatively for both occluded and non-occluded objects. Additionally, MonoMAE learns generalizable representations that can work well in new domains.

Large-vocabulary object detectors (LVDs) aim to detect objects of many categories, which learn super objectness features and can locate objects accurately while applied to various downstream data. However, LVDs often struggle in recognizing the located objects due to domain discrepancy in data distribution and object vocabulary. At the other end, recent vision-language foundation models such as CLIP demonstrate superior open-vocabulary recognition capability. This paper presents KGD, a Knowledge Graph Distillation technique that exploits the implicit knowledge graphs (KG) in CLIP for effectively adapting LVDs to various downstream domains. KGD consists of two consecutive stages: 1) KG extraction that employs CLIP to encode downstream domain data as nodes and their feature distances as edges, constructing KG that inherits the rich semantic relations in CLIP explicitly; and 2) KG encapsulation that transfers the extracted KG into LVDs to enable accurate cross-domain object classification. In addition, KGD can extract both visual and textual KG independently, providing complementary vision and language knowledge for object localization and object classification in detection tasks over various downstream domains. Experiments over multiple widely adopted detection benchmarks show that KGD outperforms the state-of-the-art consistently by large margins. Codes will be released.

Recent Large Vision Language Models (LVLMs) present remarkable zero-shot conversational and reasoning capabilities given multimodal queries. Nevertheless, they suffer from object hallucination, a phenomenon where LVLMs are prone to generate textual responses not factually aligned with image inputs. Our pilot study reveals that object hallucination is closely tied with Rotary Position Encoding (RoPE), a widely adopted positional dependency modeling design in existing LVLMs. Due to the long-term decay in RoPE, LVLMs tend to hallucinate more when relevant visual cues are distant from instruction tokens in the multimodal input sequence, Additionally, we observe a similar effect when reversing the sequential order of visual tokens during multimodal alignment. Our tests indicate that long-term decay in RoPE poses challenges to LVLMs while capturing visual-instruction interactions across long distances. We propose Concentric Causal Attention (CCA), a simple yet effective positional alignment strategy that mitigates the impact of RoPE long-term decay in LVLMs by naturally reducing relative distance between visual and instruction tokens. With CCA, visual tokens can better interact with instruction tokens, thereby enhancing model's perception capability and alleviating object hallucination. Without bells and whistles, our positional alignment method surpasses existing hallucination mitigation strategies by large margins on multiple object hallucination benchmarks.

Test-time prompt tuning, which learns prompts online with unlabelled test samples during the inference stage, has demonstrated great potential by learning effective prompts on-the-fly without requiring any task-specific annotations. However, its performance often degrades clearly along the tuning process when the prompts are continuously updated with the test data flow, and the degradation becomes more severe when the domain of test samples changes continuously. We propose HisTPT, a Historical Test-time Prompt Tuning technique that memorizes the useful knowledge of the learnt test samples and enables robust test-time prompt tuning with the memorized knowledge. HisTPT introduces three types of knowledge banks, namely, local knowledge bank, hard-sample knowledge bank, and global knowledge bank, each of which works with different mechanisms for effective knowledge memorization and test-time prompt optimization. In addition, HisTPT features an adaptive knowledge retrieval mechanism that regularizes the prediction of each test sample by adaptively retrieving the memorized knowledge. Extensive experiments show that HisTPT achieves superior prompt tuning performance consistently while handling different visual recognition tasks (e.g., image classification, semantic segmentation, and object detection) and test samples from continuously changing domains.

Masked Autoencoder (MAE) has demonstrated superb pre-training efficiency for vision Transformer, thanks to its partial input paradigm and high mask ratio (0.75). However, MAE often suffers from severe performance drop under higher mask ratios, which hinders its potential toward larger-scale vision Transformers. In this work, we identify that the performance drop is largely attributed to the over-dominance of difficult reconstruction targets, as higher mask ratios lead to more sparse visible patches and fewer visual clues for reconstruction. To mitigate this issue, we design Efficient MAE that introduces a novel Difficulty-Flatten Loss and a decoder masking strategy, enabling a higher mask ratio for more efficient pre-training. The Difficulty-Flatten Loss provides balanced supervision on reconstruction targets of different difficulties, mitigating the performance drop under higher mask ratios effectively. Additionally, the decoder masking strategy discards the most difficult reconstruction targets, which further alleviates the optimization difficulty and accelerates the pre-training clearly. Our proposed Efficient MAE introduces 27% and 30% pre-training runtime accelerations for the ViT-Large and ViT-Huge models, provides valuable insights into MAE’s optimization, and paves the way for larger-scale vision Transformer pre-training. Code and pre-trained models will be released.

With the prevalent use of LiDAR sensors in autonomous driving, 3D point cloud object tracking has received increasing attention. In a point cloud sequence, 3D object tracking aims to predict the location and orientation of an object in consecutive frames. Motivated by the success of transformers, we propose Point Tracking TRansformer (PTTR), which efficiently predicts high-quality 3D tracking results in a coarse-to-fine manner with the help of transformer operations. PTTR consists of three novel designs. 1) Instead of random sampling, we design Relation-Aware Sampling to preserve relevant points to the given template during subsampling. 2) We propose a Point Relation Transformer for effective feature aggregation and feature matching between the template and search region. 3) Based on the coarse tracking results, we employ a novel Prediction Refinement Module to obtain the final refined prediction through local feature pooling. In addition, motivated by the favorable properties of the Bird's-Eye View (BEV) of point clouds in capturing object motion, we further design a more advanced framework named PTTR++, which incorporates both the point-wise view and BEV representation to exploit their complementary effect in generating high-quality tracking results. PTTR++ substantially boosts the tracking performance on top of PTTR with low computational overhead. Extensive experiments over multiple datasets show that our proposed approaches achieve superior 3D tracking accuracy and efficiency.

In the past decade, deep neural networks have achieved significant progress in point cloud learning. However, collecting large-scale precisely-annotated point clouds is extremely laborious and expensive, which hinders the scalability of existing point cloud datasets and poses a bottleneck for efficient exploration of point cloud data in various tasks and applications. Label-efficient learning offers a promising solution by enabling effective deep network training with much-reduced annotation efforts. This paper presents the first comprehensive survey of label-efficient learning of point clouds. We address three critical questions in this emerging research field: i) the importance and urgency of label-efficient learning in point cloud processing, ii) the subfields it encompasses, and iii) the progress achieved in this area. To this end, we propose a taxonomy that organizes label-efficient learning methods based on the data prerequisites provided by different types of labels. We categorize four typical label-efficient learning approaches that significantly reduce point cloud annotation efforts: data augmentation, domain transfer learning, weakly-supervised learning, and pretrained foundation models. For each approach, we outline the problem setup and provide an extensive literature review that showcases relevant progress and challenges. Finally, we share our views on the current research challenges and potential future directions.

One-shot skeleton action recognition, which aims to learn a skeleton action recognition model with a single training sample, has attracted increasing interest due to the challenge of collecting and annotating large-scale skeleton action data. However, most existing studies match skeleton sequences by comparing their feature vectors directly which neglects spatial structures and temporal orders of skeleton data. This paper presents a novel one-shot skeleton action recognition technique that handles skeleton action recognition via multi-scale spatial-temporal feature matching. We represent skeleton data at multiple spatial and temporal scales and achieve optimal feature matching from two perspectives. The first is multi-scale matching which captures the scale-wise semantic relevance of skeleton data at multiple spatial and temporal scales simultaneously. The second is cross-scale matching which handles different motion magnitudes and speeds by capturing sample-wise relevance across multiple scales. Extensive experiments over three large-scale datasets (NTU RGB+D, NTU RGB+D 120, and PKU-MMD) show that our method achieves superior one-shot skeleton action recognition, and outperforms SOTA consistently by large margins.

3D Skeleton-based human action recognition has attracted increasing attention in recent years. Most of the existing work focuses on supervised learning which requires a large number of labeled action sequences that are often expensive and time-consuming to annotate. In this paper, we address self-supervised 3D action representation learning for skeleton-based action recognition. We investigate self-supervised representation learning and design a novel skeleton cloud colorization technique that is capable of learning spatial and temporal skeleton representations from unlabeled skeleton sequence data. We represent a skeleton action sequence as a 3D skeleton cloud and colorize each point in the cloud according to its temporal and spatial orders in the original (unannotated) skeleton sequence. Leveraging the colorized skeleton point cloud, we design an auto-encoder framework that can learn spatial-temporal features from the artificial color labels of skeleton joints effectively. Specifically, we design a two-steam pretraining network that leverages fine-grained and coarse-grained colorization to learn multi-scale spatial-temporal features. In addition, we design a Masked Skeleton Cloud Repainting task that can pretrain the designed auto-encoder framework to learn informative representations. We evaluate our skeleton cloud colorization approach with linear classifiers trained under different configurations, including unsupervised, semi-supervised, fully-supervised, and transfer learning settings. Extensive experiments on NTU RGB+D, NTU RGB+D 120, PKU-MMD, NW-UCLA, and UWA3D datasets show that the proposed method outperforms existing unsupervised and semi-supervised 3D action recognition methods by large margins and achieves competitive performance in supervised 3D action recognition as well.

Most visual recognition studies rely heavily on crowd-labelled data in deep neural networks (DNNs) training, and they usually train a DNN for each single visual recognition task, leading to a laborious and time-consuming visual recognition paradigm. To address the two challenges, Vision-Language Models (VLMs) have been intensively investigated recently, which learns rich vision-language correlation from web-scale image-text pairs that are almost infinitely available on the Internet and enables zero-shot predictions on various visual recognition tasks with a single VLM. This paper provides a systematic review of visual language models for various visual recognition tasks, including: (1) the background that introduces the development of visual recognition paradigms; (2) the foundations of VLM that summarize the widely-adopted network architectures, pre-training objectives, and downstream tasks; (3) the widely-adopted datasets in VLM pre-training and evaluations; (4) the review and categorization of existing VLM pre-training methods, VLM transfer learning methods, and VLM knowledge distillation methods; (5) the benchmarking, analysis and discussion of the reviewed methods; (6) several research challenges and potential research directions that could be pursued in the future VLM studies for visual recognition.

Pixel-wise contrastive learning recently offers a new training paradigm in semantic segmentation by directly shaping the pixel embedding space. Compared with pixel-pixel contrast that often requires large memory and high computation cost, pixel-prototype contrast exploits the semantic correlations among pixels in a more efficient way by pulling positive pixel-prototype pairs close and pushing negative pairs apart. However, most existing work treats pixels as anchors to form contrast, either failing to capture the intra-class variance or introducing extra computational overhead. In this work, we propose Prototype-Anchor Contrast (ProAC), a novel prototypical contrastive learning paradigm that strengthens pixel-prototype associations in a simple yet effective fashion. First, ProAC pre-defines class prototypes (serving as cluster centroids) by exploiting the uniformity on the hypersphere in the feature space and thus requires no prototype updating during network optimization, which greatly simplifies the network training process. Second, by treating prototypes as anchors, ProAC builds a novel prototype-to-pixel learning path, where a large amount of negative pixels can naturally be generated to describe rich semantic information without relying on auxiliary sample augmentation techniques. Finally, as a plug-and-play regularization term, ProAC can be attached to most existing segmentation models and assist the network optimization by directly shaping the pixel embedding space. Extensive experiments on different benchmarks show that our ProAC brings an mIoU increase from 1.4% to 2.0% for fully-supervised models and from 0.9% to 6.0% for domain-adaptive models, respectively. It also leads to a gain of mIoU, ranging from 1.8% to 2.7% in more challenging cases, including different resolutions, diverse illuminations and masked scenarios.

Domain adaptive LiDAR point cloud segmentation aims to learn a target segmentation model from labeled source point clouds and unlabelled target point clouds, which has recently attracted increasing attention due to various challenges in point cloud annotation. However, its performance is still very constrained as most existing studies did not well capture data-specific characteristics of LiDAR point clouds. Inspired by the observation that the domain discrepancy of LiDAR point clouds is highly correlated with point density, we design a density-aware self-training (DAST) technique that introduces point density into the self-training framework for domain adaptive point cloud segmentation. DAST consists of two novel and complementary designs. The first is density-aware pseudo labelling that introduces point density for accurate pseudo labelling of target data and effective self-supervised network retraining. The second is density-aware consistency regularization that encourages to learn density-invariant representations by enforcing target predictions to be consistent across points of different densities. Extensive experiments over multiple large-scale public datasets show that DAST achieves superior domain adaptation performance as compared with the state-of-the-art.

Cross-domain semantic segmentation, which aims to address the distribution shift while adapting from a labeled source domain to an unlabeled target domain, has achieved great progress in recent years. However, most existing work adopts a source-to-target adaptation path, which often suffers from clear class mismatching or class imbalance issues. We design PBAL, a prototypical bidirectional adaptation and learning technique that introduces bidirectional prototype learning and prototypical self-training for optimal inter-domain alignment and adaptation. We perform bidirectional alignments in a complementary and cooperative manner which balances both dominant and tail categories as well as easy and hard samples effectively. In addition, We derive prototypes efficiently from a source-trained classifier, which enables class-aware adaptation as well as synchronous prototype updating and network optimization. Further, we re-examine self-training and introduce prototypical contrast above it which greatly improves inter-domain alignment by promoting better intra-class compactness and inter-class separability in the feature space. Extensive experiments over two widely studied benchmarks show that the proposed PBAL achieves superior domain adaptation performance as compared with the state-of-the-art.

Domain adaptive LiDAR point cloud segmentation aims to learn an effective target segmentation model from labelled source data and unlabelled target data, which has attracted increasing attention in recent years due to the difficulty in point-cloud annotation. It remains a very open research challenge as point clouds of different domains often have clear distribution discrepancies with variations in LiDAR sensor configurations, environmental conditions, occlusions, etc. We design a simple yet effective spatial consistency training framework that can learn superior domain-invariant feature representations from unlabelled target point clouds. The framework exploits three types of spatial consistency, namely, geometric-transform consistency, sparsity consistency, and mixing consistency which capture the semantic invariance of point clouds with respect to viewpoint changes, sparsity changes, and local context changes, respectively. With a concise mean teacher learning strategy, our experiments show that the proposed spatial consistency training outperforms the state-of-the-art significantly and consistently across multiple public benchmarks.

Generative neural radiance fields (NeRF) bring image generation into the 3D era, which have delivered impressive generation quality and 3D consistency, especially in the face generation domain. Upon pre-trained generative NeRF, 3D-aware image editing has been explored and achieved promising performance via manipulating semantic maps or attributes. However, a more flexible editing interface, text, remains under-explored in the context of 3D-aware image editing. In this work, we leverage the Contrastive Language-Image Pre-training (CLIP) model to achieve 3D-aware image editing in pre-trained generative NeRF models given a target text prompt. To achieve accurate and controllable geometry editing, we propose MorphNeRF, a learnable morphing network that morphs the 3D geometry of images toward the target descriptions via generative NeRF. Different from prior studies that achieve image editing by manipulating latent codes or directly finetuning pre-trained models, morphing the geometry can better preserve the texture of the source image and facilitate the control of editing strength by adjusting the weight of morphing maps explicitly. Extensive experiments and comparisons show that the proposed MorphNeRF achieves superior image editing performance.

Training deep neural networks (DNNs) with noisy labels often leads to poorly generalized models as DNNs tend to memorize the noisy labels in training. Various strategies have been developed for improving sample selection precision and mitigating the noisy label memorization issue. However, most existing works adopt a class-dependent softmax classifier that is vulnerable to noisy labels by entangling the classification of multi-class features. This paper presents a class-independent regularization (CIR) method that can effectively alleviate the negative impact of noisy labels in DNN training. CIR regularizes the class-dependent softmax classifier by introducing multi-binary classifiers each of which takes care of one class only. Thanks to its class-independent nature, CIR is tolerant to noisy labels as misclassification by one binary classifier does not affect others. For effective training of CIR, we design a heterogeneous adaptive co-teaching strategy that forces the class-independent and class-dependent classifiers to focus on sample selection and image classification, respectively, in a cooperative manner. Extensive experiments show that CIR achieves superior performance consistently across multiple benchmarks with both synthetic and real images. Code is available at https://github.com/RumengYi/CIR.

Domain adaptive panoptic segmentation aims to miti-gate data annotation challenge by leveraging off-the-shelf annotated data in one or multiple related source domains. However, existing studies employ two separate networks for instance segmentation and semantic segmentation which lead to excessive network parameters as well as complicated and computationally intensive training and inference processes. We design UniDAformer, a unified domain adaptive panoptic segmentation transformer that is simple but can achieve domain adaptive instance segmentation and semantic segmentation simultaneously within a single network. UniDAformer introduces Hierarchical Mask Calibration (HMC) that rectifies inaccurate predictions at the level of regions, superpixels and pixels via online self-training on the fly. It has three unique features: 1) it enables unified domain adaptive panoptic adaptation; 2) it mitigates false predictions and improves domain adaptive panoptic segmentation effectively; 3) it is end-to-end trainable with a much simpler training and inference pipeline. Exten-sive experiments over multiple public benchmarks show that UniDAformer achieves superior domain adaptive panoptic segmentation as compared with the state-of-the-art.

Generative Adversarial Networks (GANs) rely heavily on large-scale training data for training high-quality image generation models. With limited training data, the GAN discriminator often suffers from severe overfitting which directly leads to degraded generation especially in generation diversity. Inspired by the recent advances in knowledge distillation (KD), we propose KD-DLGAN, a knowledge-distillation based generation framework that introduces pre-trained vision-language models for training effective data-limited generation models. KD-DLGAN consists of two innovative designs. The first is aggregated generative KD that mitigates the discriminator overfitting by challenging the discriminator with harder learning tasks and distilling more generalizable knowledge from the pre-trained models. The second is correlated generative KD that improves the generation diversity by distilling and preserving the diverse image-text correlation within the pre-trained models. Extensive experiments over multiple benchmarks show that KD-DLGAN achieves superior image generation with limited training data. In addition, KD-DLGAN complements the state-of-the-art with consistent and substantial performance gains. Note that codes will be released.

Contrastive learning has recently demonstrated great potential for unsupervised pre-training in 3D scene understanding tasks. However, most existing work randomly selects point features as anchors while building contrast, leading to a clear bias toward background points that often dominate in 3D scenes. Also, object awareness and foreground-to-background discrimination are neglected, making contrastive learning less effective. To tackle these issues, we propose a general foreground-aware feature contrast (FAC) framework to learn more effective point cloud representations in pre-training. FAC consists of two novel contrast designs to construct more effective and informative contrast pairs. The first is building positive pairs within the same foreground segment where points tend to have the same semantics. The second is that we prevent over-discrimination between 3D segments/objects and encourage foreground-to-background distinctions at the segment level with adaptive feature learning in a Siamese correspondence network, which adaptively learns feature correlations within and across point cloud views effectively. Visualization with point activation maps shows that our contrast pairs capture clear correspondences among fore-ground regions during pre-training. Quantitative experiments also show that FAC achieves superior knowledge transfer and data efficiency in various downstream 3D semantic segmentation and object detection tasks. All codes, data, and models are available.

3D style transfer aims to render stylized novel views of a 3D scene with multiview consistency. However, most existing work suffers from a three-way dilemma over accurate geometry reconstruction, high-quality stylization, and being generalizable to arbitrary new styles. We propose StyleRF (Style Radiance Fields), an innovative 3D style transfer technique that resolves the three-way dilemma by performing style transformation within the feature space of a radiance field. StyleRF employs an explicit grid of high-level features to represent 3D scenes, with which highfidelity geometry can be reliably restored via volume rendering. In addition, it transforms the grid features according to the reference style which directly leads to high-quality zero-shot style transfer. StyleRF consists of two innovative designs. The first is sampling-invariant content transformation that makes the transformation invariant to the holistic statistics of the sampled 3D points and accordingly ensures multi-view consistency. The second is deferred style transformation of 2D feature maps which is equivalent to the transformation of 3D points but greatly reduces memory footprint without degrading multi-view consistency. Extensive experiments show that StyleRF achieves superior 3D stylization quality with precise geometry reconstruction and it can generalize to various new styles in a zero-shot manner. Project website: https://kunhao-liu.github.io/StyleRF/

Robust point cloud parsing under all-weather conditions is crucial to level-5 autonomy in autonomous driving. However, how to learn a universal 3D semantic segmentation (3DSS) model is largely neglected as most existing benchmarks are dominated by point clouds captured under normal weather. We introduce SemanticSTF, an adverse-weather point cloud dataset that provides dense point-level annotations and allows to study 3DSS under various adverse weather conditions. We study all-weather 3DSS modeling under two setups: 1) domain adaptive 3DSS that adapts from normal-weather data to adverse-weather data; 2) domain generalizable 3DSS that learns all-weather 3DSS models from normal-weather data. Our studies reveal the challenge while existing 3DSS methods encounter adverse-weather data, showing the great value of SemanticSTF in steering the future endeavor along this very meaningful research direction. In addition, we design a domain randomization technique that alternatively randomizes the geometry styles of point clouds and aggregates their embeddings, ultimately leading to a generalizable model that can improve 3DSS under various adverse weather effectively. The SemanticSTF and related codes are available at https://github.com/xiaoaoran/SemanticSTF.

Recent deep-learning-based compression methods have achieved superior performance compared with traditional approaches. However, deep learning models have proven to be vulnerable to backdoor attacks, where some specific trigger patterns added to the input can lead to malicious behavior of the models. In this paper, we present a novel backdoor attack with multiple triggers against learned image compression models. Motivated by the widely used discrete cosine transform (DCT) in existing compression systems and standards, we propose a frequency-based trigger injection model that adds triggers in the DCT domain. In particular, we design several attack objectives for various attacking scenarios, including: 1) attacking compression quality in terms of bit-rate and reconstruction quality; 2) attacking task-driven measures, such as downstream face recognition and semantic segmentation. Moreover, a novel simple dynamic loss is designed to balance the influence of different loss terms adaptively, which helps achieve more efficient training. Extensive experiments show that with our trained trigger injection models and simple modification of encoder parameters (of the compression model), the proposed attack can successfully inject several backdoors with corresponding triggers in a single image compression model.

The recent detection transformer (DETR) simplifies the object detection pipeline by removing hand-crafted designs and hyperparameters as employed in conventional two-stage object detectors. However, how to leverage the simple yet effective DETR architecture in domain adaptive object detection is largely neglected. Inspired by the unique DETR attention mechanisms, we design DA-DETR, a domain adaptive object detection transformer that introduces information fusion for effective transfer from a labeled source domain to an unlabeled target domain. DA-DETR introduces a novel CNN-Transformer Blender (CTBlender) that fuses the CNN features and Transformer features ingeniously for effective feature alignment and knowledge transfer across domains. Specifically, CTBlender employs the Transformer features to modulate the CNN features across multiple scales where the high-level semantic information and the low-level spatial information are fused for accurate object identification and localization. Extensive experiments show that DA-DETR achieves superior detection performance consistently across multiple widely adopted domain adaptation benchmarks.

Multi-scale features have been proven highly effective for object detection but often come with huge and even prohibitive extra computation costs, especially for the recent Transformer-based detectors. In this paper, we propose Iterative Multi-scale Feature Aggregation (IMFA) - a generic paradigm that enables efficient use of multi-scale features in Transformer-based object detectors. The core idea is to exploit sparse multi-scale features from just a few crucial locations, and it is achieved with two novel designs. First, IMFA rearranges the Transformer encoder-decoder pipeline so that the encoded features can be iteratively updated based on the detection predictions. Second, IMFA sparsely samples scale-adaptive features for refined detection from just a few keypoint locations under the guidance of prior detection predictions. As a result, the sampled multi-scale features are sparse yet still highly beneficial for object detection. Extensive experiments show that the proposed IMFA boosts the performance of multiple Transformer-based object detectors significantly yet with only slight computational overhead.

Quantizing images into discrete representations has been a fundamental problem in unified generative modeling. Predominant approaches learn the discrete representation either in a deterministic manner by selecting the best-matching token or in a stochastic manner by sampling from a predicted distribution. However, deterministic quantization suffers from severe codebook collapse and misalignment with inference stage while stochastic quantization suffers from low codebook utilization and perturbed reconstruction objective. This paper presents a regularized vector quantization framework that allows to mitigate above issues effectively by applying regularization from two perspectives. The first is a prior distribution regularization which measures the discrepancy between a prior token distribution and the predicted token distribution to avoid code-book collapse and low codebook utilization. The second is a stochastic mask regularization that introduces stochasticity during quantization to strike a good balance between inference stage misalignment and unperturbed reconstruction objective. In addition, we design a probabilistic contrastive loss which serves as a calibrated metric to further mitigate the perturbed reconstruction objective. Extensive experiments show that the proposed quantization framework outperforms prevailing vector quantization methods consistently across different generative models including auto-regressive models and diffusion models.

Black-box unsupervised domain adaptation (UDA) learns with source predictions of target data without accessing either source data or source models during training, and it has clear superiority in data privacy and flexibility in target network selection. However, the source predictions of target data are often noisy and training with them is prone to learning collapses. We propose BiMem, a bidirectional memorization mechanism that learns to remember useful and representative information to correct noisy pseudo labels on the fly, leading to robust black-box UDA that can generalize across different visual recognition tasks. BiMem constructs three types of memory, including sensory memory, short-term memory, and long-term memory, which interact in a bi-directional manner for comprehensive and robust memorization of learnt features. It includes a forward memorization flow that identifies and stores useful features and a backward calibration flow that rectifies features’ pseudo labels progressively. Extensive experiments show that BiMem achieves superior domain adaptation performance consistently across various visual recognition tasks such as image classification, semantic segmentation and object detection.

Domain generalization (DG) aims to learn domain-generalizable models from one or multiple source domains that can perform well in unseen target domains. Despite its recent progress, most existing work suffers from the misalignment between the difficulty level of training samples and the capability of contemporarily trained models, leading to over-fitting or under-fitting in the trained generalization model. We design MoDify, a Momentum Difficulty framework that tackles the misalignment by balancing the seesaw between the model’s capability and the samples’ difficulties along the training process. MoDify consists of two novel designs that collaborate to fight against the misalignment while learning domain-generalizable models. The first is MoDify-based Data Augmentation which exploits an RGB Shuffle technique to generate difficulty-aware training samples on the fly. The second is MoDify-based Network Optimization which dynamically schedules the training samples for balanced and smooth learning with appropriate difficulty. Without bells and whistles, a simple implementation of MoDify achieves superior performance across multiple benchmarks. In addition, MoDify can complement existing methods as a plug-in, and it is generic and can work for different visual recognition tasks.

Neural Radiance Field (NeRF) has shown impressive performance in novel view synthesis via implicit scene representation. However, it usually suffers from poor scalability as requiring densely sampled images for each new scene. Several studies have attempted to mitigate this problem by integrating Multi-View Stereo (MVS) technique into NeRF while they still entail a cumbersome fine-tuning process for new scenes. Notably, the rendering quality will drop severely without this fine-tuning process and the errors mainly appear around the high-frequency features. In the light of this observation, we design WaveNeRF, which integrates wavelet frequency decomposition into MVS and NeRF to achieve generalizable yet high-quality synthesis without any per-scene optimization. To preserve high-frequency information when generating 3D feature volumes, WaveNeRF builds Multi-View Stereo in the Wavelet domain by integrating the discrete wavelet transform into the classical cascade MVS, which disentangles high-frequency information explicitly. With that, disentangled frequency features can be injected into classic NeRF via a novel hybrid neural renderer to yield faithful high-frequency details, and an intuitive frequency-guided sampling strategy can be designed to suppress artifacts around high-frequency regions. Extensive experiments over three widely studied benchmarks show that WaveNeRF achieves superior generalizable radiance field modeling when only given three images as input.

Pose-free neural radiance fields (NeRF) aim to train NeRF with unposed multi-view images and it has achieved very impressive success in recent years. Most existing works share the pipeline of training a coarse pose estimator with rendered images at first, followed by a joint optimization of estimated poses and neural radiance field. However, as the pose estimator is trained with only rendered images, the pose estimation is usually biased or inaccurate for real images due to the domain gap between real images and rendered images, leading to poor robustness for the pose estimation of real images and further local minima in joint optimization. We design IR-NeRF, an innovative pose-free NeRF that introduces implicit pose regularization to refine pose estimator with unposed real images and improve the robustness of the pose estimation for real images. With a collection of 2D images of a specific scene, IR-NeRF constructs a scene codebook that stores scene features and captures the scene-specific pose distribution implicitly as priors. Thus, the robustness of pose estimation can be promoted with the scene priors according to the rationale that a 2D real image can be well reconstructed from the scene codebook only when its estimated pose lies within the pose distribution. Extensive experiments show that IR-NeRF achieves superior novel view synthesis and outperforms the state-of-the-art consistently across multiple synthetic and real datasets.

Vision-Language Pre-training has demonstrated its remarkable zero-shot recognition ability and potential to learn generalizable visual representations from languagesupervision. Taking a step ahead, language-supervised semantic segmentation enables spatial localization of textual inputs by learning pixel grouping solely from image-text pairs. Nevertheless, the state-of-the-art suffers from a clear semantic gap between visual and textual modalities: plenty of visual concepts appeared in images are missing in their paired captions. Such semantic misalignment circulates in pre-training, leading to inferior zero-shot performance in dense predictions due to insufficient visual concepts captured in textual representations. To close such semantic gap, we propose Concept Curation (CoCu), a pipeline that leverages CLIP to compensate for the missing semantics. For each image-text pair, we establish a concept archive that maintains potential visually-matched concepts with our proposed vision-driven expansion and text-to-vision-guided ranking. Relevant concepts can thus be identified via cluster-guided sampling and fed into pre-training, thereby bridging the gap between visual and textual semantics. Extensive experiments over a broad suite of 8 segmentation benchmarks show that CoCu achieves superb zero-shot transfer performance and greatly boosts language-supervised segmentation baseline by a large margin, suggesting the value of closing semantic gap in pre-training data.

Open-vocabulary segmentation of 3D scenes is a fundamental function of human perception and thus a crucial objective in computer vision research. However, this task is heavily impeded by the lack of large-scale and diverse 3D open-vocabulary segmentation datasets for training robust and generalizable models. Distilling knowledge from pre-trained 2D open-vocabulary segmentation models helps but it compromises the open-vocabulary feature as the 2D models are mostly finetuned with close-vocabulary datasets. We tackle the challenges in 3D open-vocabulary segmentation by exploiting pre-trained foundation models CLIP and DINO in a weakly supervised manner. Specifically, given only the open-vocabulary text descriptions of the objects in a scene, we distill the open-vocabulary multimodal knowledge and object reasoning capability of CLIP and DINO into a neural radiance field (NeRF), which effectively lifts 2D features into view-consistent 3D segmentation. A notable aspect of our approach is that it does not require any manual segmentation annotations for either the foundation models or the distillation process. Extensive experiments show that our method even outperforms fully supervised models trained with segmentation annotations in certain scenes, suggesting that 3D open-vocabulary segmentation can be effectively learned from 2D images and text-image pairs. Code is available at https://github.com/Kunhao-Liu/3D-OVS.

High-definition (HD) vectorized map is essential for autonomous driving, providing detailed and precise environmental information for advanced perception and planning. However, current map vectorization methods often exhibit deviations, and the existing evaluation metric for map vectorization lacks sufficient sensitivity to detect these deviations. To address these limitations, we propose integrating the philosophy of rasterization into map vectorization. Specifically, we introduce a new rasterization-based evaluation metric, which has superior sensitivity and is better suited to real-world autonomous driving scenarios. Furthermore, we propose MapVR (Map Vectorization via Rasterization), a novel framework that applies differentiable rasterization to vectorized outputs and then performs precise and geometry-aware supervision on rasterized HD maps. Notably, MapVR designs tailored rasterization strategies for various geometric shapes, enabling effective adaptation to a wide range of map elements. Experiments show that incorporating rasterization into map vectorization greatly enhances performance with no extra computational cost during inference, leading to more accurate map perception and ultimately promoting safer autonomous driving. Codes are available at https://github.com/ZhangGongjie/MapVR. A standalone map vectorization evaluation toolkit is available at https://github.com/jiahaoLjh/MapVectorizationEvalToolkit.

3D object detection using point clouds has attracted increasing attention due to its wide applications in autonomous driving and robotics. However, most existing studies focus on single point cloud frames without harnessing the temporal information in point cloud sequences. In this paper, we design TransPillars, a novel transformer-based feature aggregation technique that exploits temporal features of consecutive point cloud frames for multi-frame 3D object detection. TransPillars aggregates spatial-temporal point cloud features from two perspectives. First, it fuses voxel-level features directly from multi-frame feature maps instead of pooled instance features to preserve instance details with contextual information that are essential to accurate object localization. Second, it introduces a hierarchical coarse-to-fine strategy to fuse multi-scale features progressively to effectively capture the motion of moving objects and guide the aggregation of fine features. Besides, a variant of deformable transformer is introduced to improve the effectiveness of cross-frame feature matching. Extensive experiments show that our proposed TransPillars achieves state-of-art performance as compared to existing multi-frame detection approaches.

Predicting human motion from historical pose sequence is crucial for a machine to succeed in intelligent interactions with humans. One aspect that has been obviated so far, is the fact that how we represent the skeletal pose has a critical impact on the prediction results. Yet there is no effort that investigates across different pose representation schemes. We conduct an indepth study on various pose representations with a focus on their effects on the motion prediction task. Moreover, recent approaches build upon off-the-shelf RNN units for motion prediction. These approaches process input pose sequence sequentially and inherently have difficulties in capturing long-term dependencies. In this paper, we propose a novel RNN architecture termed AHMR (Attentive Hierarchical Motion Recurrent network) for motion prediction which simultaneously models local motion contexts and a global context. We further explore a geodesic loss and a forward kinematics loss for the motion prediction task, which have more geometric significance than the widely employed L2 loss. Interestingly, we applied our method to a range of articulate objects including human, fish, and mouse. Empirical results show that our approach outperforms the state-of-the-art methods in short-term prediction and achieves much enhanced long-term prediction proficiency, such as retaining natural human-like motions over 50 seconds predictions. Our codes are released.

Point cloud data have been widely explored due to its superior accuracy and robustness under various adverse situations. Meanwhile, deep neural networks (DNNs) have achieved very impressive success in various applications such as surveillance and autonomous driving. The convergence of point cloud and DNNs has led to many deep point cloud models, largely trained under the supervision of large-scale and densely-labelled point cloud data. Unsupervised point cloud representation learning, which aims to learn general and useful point cloud representations from unlabelled point cloud data, has recently attracted increasing attention due to the constraint in large-scale point cloud labelling. This paper provides a comprehensive review of unsupervised point cloud representation learning using DNNs. It first describes the motivation, general pipelines as well as terminologies of the recent studies. Relevant background including widely adopted point cloud datasets and DNN architectures is then briefly presented. This is followed by an extensive discussion of existing unsupervised point cloud representation learning methods according to their technical approaches. We also quantitatively benchmark and discuss the reviewed methods over multiple widely adopted point cloud datasets. Finally, we share our humble opinion about several challenges and problems that could be pursued in the future research in unsupervised point cloud representation learning.

Leveraging the advances of natural language processing, most recent scene text recognizers adopt an encoder-decoder architecture where text images are first converted to representative features and then a sequence of characters via 'sequential decoding'. However, scene text images suffer from rich noises of different sources such as complex background and geometric distortions which often confuse the decoder and lead to incorrect alignment of visual features at noisy decoding time steps. This paper presents I2C2W, a novel scene text recognition technique that is tolerant to geometric and photometric degradation by decomposing scene text recognition into two inter-connected tasks. The first task focuses on image-to-character (I2C) mapping which detects a set of character candidates from images based on different alignments of visual features in an non-sequential way. The second task tackles character-to-word (C2W) mapping which recognizes scene text by decoding words from the detected character candidates. The direct learning from character semantics (instead of noisy image features) corrects falsely detected character candidates effectively which improves the final text recognition accuracy greatly. Extensive experiments over nine public datasets show that the proposed I2C2W outperforms the state-of-the-art by large margins for challenging scene text datasets with various curvature and perspective distortions. It also achieves very competitive recognition performance over multiple normal scene text datasets.

As information exists in various modalities in real world, effective interaction and fusion among multimodal information plays a key role for the creation and perception of multimodal data in computer vision and deep learning research. With superb power in modeling the interaction among multimodal information, multimodal image synthesis and editing has become a hot research topic in recent years. Instead of providing explicit guidance for network training, multimodal guidance offers intuitive and flexible means for image synthesis and editing. On the other hand, this field is also facing several challenges in alignment of multimodal features, synthesis of high-resolution images, faithful evaluation metrics, etc. In this survey, we comprehensively contextualize the advance of the recent multimodal image synthesis and editing and formulate taxonomies according to data modalities and model types. We start with an introduction to different guidance modalities in image synthesis and editing, and then describe multimodal image synthesis and editing approaches extensively according to their model types. After that, we describe benchmark datasets and evaluation metrics as well as corresponding experimental results. Finally, we provide insights about the current research challenges and possible directions for future research.

Few-shot object detection has been extensively investigated by incorporating meta-learning into region-based detection frameworks. Despite its success, the said paradigm is still constrained by several factors, such as (i) low-quality region proposals for novel classes and (ii) negligence of the inter-class correlation among different classes. Such limitations hinder the generalization of base-class knowledge for the detection of novel-class objects. In this work, we design Meta-DETR, which (i) is the first image-level few-shot detector, and (ii) introduces a novel inter-class correlational meta-learning strategy to capture and leverage the correlation among different classes for robust and accurate few-shot object detection. Meta-DETR works entirely at image level without any region proposals, which circumvents the constraint of inaccurate proposals in prevalent few-shot detection frameworks. In addition, the introduced correlational meta-learning enables Meta-DETR to simultaneously attend to multiple support classes within a single feedforward, which allows to capture the inter-class correlation among different classes, thus significantly reducing the misclassification over similar classes and enhancing knowledge generalization to novel classes. Experiments over multiple few-shot object detection benchmarks show that the proposed Meta-DETR outperforms state-of-the-art methods by large margins. The implementation codes are publicly available at https://github.com/ZhangGongjie/Meta-DETR.

Audio-driven talking face generation, which aims to synthesize talking faces with realistic facial animations (including accurate lip movements, vivid facial expression details and natural head poses) corresponding to the audio, has achieved rapid progress in recent years. However, most existing work focuses on generating lip movements only without handling the closely correlated facial expressions, which degrades the realism of the generated faces greatly. This paper presents DIRFA, a novel method that can generate talking faces with diverse yet realistic facial animations from the same driving audio. To accommodate fair variation of plausible facial animations for the same audio, we design a transformer-based probabilistic mapping network that can model the variational facial animation distribution conditioned upon the input audio and autoregressively convert the audio signals into a facial animation sequence. In addition, we introduce a temporally-biased mask into the mapping network, which allows to model the temporal dependency of facial animations and produce temporally smooth facial animation sequence. With the generated facial animation sequence and a source image, photo-realistic talking faces can be synthesized with a generic generation network. Extensive experiments show that DIRFA can generate talking faces with realistic facial animations effectively.

Unsupervised cross-domain Facial Expression Recognition (FER) aims to transfer the knowledge from a labeled source domain to an unlabeled target domain. Existing methods strive to reduce the discrepancy between source and target domain, but cannot effectively explore the abundant semantic information of the target domain due to the absence of target labels. To this end, we propose a novel framework via Contrastive Warm up and Complexity-aware Self-Training (namely CWCST), which facilitates source knowledge transfer and target semantic learning jointly. Specifically, we formulate a contrastive warm up strategy via features, momentum features, and learnable category centers to concurrently learn discriminative representations and narrow the domain gap, which benefits domain adaptation by generating more accurate target pseudo labels. Moreover, to deal with the inevitable noise in pseudo labels, we develop complexity-aware self-training with a label selection module based on prediction entropy, which iteratively generates pseudo labels and adaptively chooses the reliable ones for training, ultimately yielding effective target semantics exploration. Furthermore, by jointly using the two mentioned components, our framework enables to effectively utilize the source knowledge and target semantic information by source-target co- training. In addition, our framework can be easily incorporated into other baselines with consistent performance improvements. Extensive experimental results on seven databases show the superior performance of the proposed method against various baselines.

Facial expression editing has attracted increasing attention with the advance of deep neural networks in recent years. However, most existing methods suffer from compromised editing fidelity and limited usability as they either ignore pose variations (unrealistic editing) or require paired training data (not easy to collect) for pose controls. This paper presents POCE, an innovative pose-controllable expression editing network that can generate realistic facial expressions and head poses simultaneously with just unpaired training images. POCE achieves the more accessible and realistic pose-controllable expression editing by mapping face images into UV space, where facial expressions and head poses can be disentangled and edited separately. POCE has two novel designs. The first is self-supervised UV completion that allows to complete UV maps sampled under different head poses, which often suffer from self-occlusions and missing facial texture. The second is weakly-supervised UV editing that allows to generate new facial expressions with minimal modification of facial identity, where the synthesized expression could be controlled by either an expression label or directly transplanted from a reference UV map via feature transfer. Extensive experiments show that POCE can learn from unpaired face images effectively, and the learned model can generate realistic and high-fidelity facial expressions under various new poses.

Person re-IDentification (re-ID) under various occlusions has been a long-standing challenge as person images with different types of occlusions often suffer from misalignment in image matching and ranking. Most existing methods tackle this challenge by aligning spatial features of body parts according to external semantic cues or feature similarities but this alignment approach is complicated and sensitive to noises. We design DRL-Net, a disentangled representation learning network that handles occluded re-ID without requiring strict person image alignment or any additional supervision. Leveraging transformer architectures, DRL-Net achieves alignment-free re-ID via global reasoning of local features of occluded person images. It measures image similarity by automatically disentangling the representation of undefined semantic components, e.g., human body parts or obstacles, under the guidance of semantic preference object queries in the transformer. In addition, we design a decorrelation constraint in the transformer decoder and impose it over object queries for better focus on different semantic components. To better eliminate interference from occlusions, we design a contrast feature learning technique (CFL) for better separation of occlusion features and discriminative ID features. Extensive experiments over occluded and holistic re-ID benchmarks show that the DRL-Net achieves superior re-ID performance consistently and outperforms the state-offi-the-art by large margins for occluded re-ID dataset.

Training effective Generative Adversarial Networks (GANs) requires large amounts of training data, without which the trained models are usually sub-optimal with discriminator over-fitting. Several prior studies address this issue by expanding the distribution of the limited training data via massive and hand-crafted data augmentation. We handle data-limited image generation from a very different perspective. Specifically, we design GenCo, a Generative Co-training network that mitigates the discriminator over-fitting issue by introducing multiple complementary discriminators that provide diverse supervision from multiple distinctive views in training. We instantiate the idea of GenCo in two ways. The first way is Weight-Discrepancy Co-training (WeCo) which co-trains multiple distinctive discriminators by diversifying their parameters. The second way is Data-Discrepancy Co-training (DaCo) which achieves co-training by feeding discriminators with different views of the input images. Extensive experiments over multiple benchmarks show that GenCo achieves superior generation with limited training data. In addition, GenCo also complements the augmentation approach with consistent and clear performance gains when combined.

Knowledge transfer from synthetic to real data has been widely studied to mitigate data annotation constraints in various computer vision tasks such as semantic segmentation. However, the study focused on 2D images and its counterpart in 3D point clouds segmentation lags far behind due to the lack of large-scale synthetic datasets and effective transfer methods. We address this issue by collecting SynLiDAR, a large-scale synthetic LiDAR dataset that contains point-wise annotated point clouds with accurate geometric shapes and comprehensive semantic classes. SynLiDAR was collected from multiple virtual environments with rich scenes and layouts which consists of over 19 billion points of 32 semantic classes. In addition, we design PCT, a novel point cloud translator that effectively mitigates the gap between synthetic and real point clouds. Specifically, we decompose the synthetic-to-real gap into an appearance component and a sparsity component and handle them separately which improves the point cloud translation greatly. We conducted extensive experiments over three transfer learning setups including data augmentation, semi-supervised domain adaptation and unsupervised domain adaptation. Extensive experiments show that SynLiDAR provides a high-quality data source for studying 3D transfer and the proposed PCT achieves superior point cloud translation consistently across the three setups. The dataset is available at https://github.com/xiaoaoran/SynLiDAR.

Instance contrast for unsupervised representation learning has achieved great success in recent years. In this work, we explore the idea of instance contrastive learning in unsupervised domain adaptation (UDA) and propose a novel Category Contrast technique (CaCo) that introduces semantic priors on top of instance discrimination for visual UDA tasks. By considering instance contrastive learning as a dictionary look-up operation, we construct a semantics-aware dictionary with samples from both source and target domains where each target sample is assigned a (pseudo) category label based on the category priors of source samples. This allows category contrastive learning (between target queries and the category-level dictionary) for category-discriminative yet domain-invariant feature representations: samples of the same category (from either source or target domain) are pulled closer while those of different categories are pushed apart simultaneously. Extensive UDA experiments in multiple visual tasks (e.g., segmentation, classification and detection) show that CaCo achieves superior performance as compared with state-of-the-art methods. The experiments also demonstrate that CaCo is complementary to existing UDA methods and gen-eralizable to other learning setups such as unsupervised model adaptation, open-/partial-set adaptation etc.

Semi-supervised semantic segmentation learns from small amounts of labelled images and large amounts of unlabelled images, which has witnessed impressive progress with the recent advance of deep neural networks. However, it often suffers from severe class-bias problem while exploring the unlabelled images, largely due to the clear pixel-wise class imbalance in the labelled images. This paper presents an unbiased subclass regularization network (USRN) that alleviates the class imbalance issue by learning class-unbiased segmentation from balanced subclass distributions. We build the balanced subclass distributions by clustering pixels of each original class into multiple subclasses of similar sizes, which provide class-balanced pseudo supervision to regularize the class-biased segmentation. In addition, we design an entropy-based gate mechanism to coordinate learning between the original classes and the clustered subclasses which facilitates subclass regularization effectively by suppressing unconfident subclass predictions. Extensive experiments over multiple public benchmarks show that USRN achieves superior performance as compared with the state-of-the-art.

State-of-the-art document dewarping techniques learn to predict 3-dimensional information of documents which are prone to errors while dealing with documents with irregular distortions or large variations in depth. This paper presents FDRNet, a Fourier Document Restoration Network that can restore documents with different distortions and improve document recognition in a reliable and simpler manner. FDRNet focuses on high-frequency components in the Fourier space that capture most structural information but are largely free of degradation in appearance. It dewarps documents by a flexible Thin-Plate Spline transformation which can handle various deformations effectively without requiring deformation annotations in training. These features allow FDRNet to learn from a small amount of simply labeled training images, and the learned model can dewarp documents with complex geometric distortion and recognize the restored texts accurately. To facilitate document restoration research, we create a benchmark dataset consisting of over one thousand camera documents with different types of geometric and photometric distortion. Extensive experiments show that FDRNet outperforms the state-of-the-art by large margins on both dewarping and text recognition tasks. In addition, FDRNet requires a small amount of simply labeled training data and is easy to deploy. The proposed dataset is available at https://sg-vilab.github.io/event/warpdoc/.

Exemplar-based image translation establishes dense correspondences between a conditional input and an exemplar (from two different domains) for leveraging detailed exemplar styles to achieve realistic image translation. Existing work builds the cross-domain correspondences implicitly by minimizing feature- wise distances across the two domains. Without explicit exploitation of domain-invariant features, this approach may not reduce the domain gap effectively which often leads to sub-optimal correspon-dences and image translation. We design a Marginal Contrastive Learning Network (MCL-Net) that explores contrastive learning to learn domain-invariant features for realistic exemplar-based image translation. Specifically, we design an innovative marginal contrastive loss that guides to establish dense correspondences explicitly. Nevertheless, building correspondence with domain-invariant semantics alone may impair the texture patterns and lead to degraded texture generation. We thus design a Self-Correlation Map (SCM) that incorporates scene structures as auxiliary information which improves the built correspondences substantially. Quantitative and qualitative experiments on multifarious image translation tasks show that the proposed method outperforms the state-of-the-art consistently.

Perceiving the similarity between images has been a long-standing and fundamental problem underlying various visual generation tasks. Predominant approaches measure the inter-image distance by computing pointwise absolute deviations, which tends to estimate the median of instance distributions and leads to blurs and artifacts in the generated images. This paper presents MoNCE, a versatile metric that introduces image contrast to learn a calibrated metric for the perception of multifaceted inter-image distances. Unlike vanilla contrast which indiscriminately pushes negative samples from the anchor regardless of their similarity, we propose to re-weight the pushing force of negative samples adaptively according to their similarity to the anchor, which facilitates the contrastive learning from informative negative samples. Since multiple patch-level contrastive objectives are involved in image distance measurement, we introduce optimal transport in MoNCE to modulate the pushing force of negative samples collaboratively across multiple contrastive objectives. Extensive experiments over multiple image translation tasks show that the proposed MoNCE outperforms various prevailing metrics substantially. The code is available at MoNCE.

Though unsupervised domain adaptation (UDA) has achieved very impressive progress recently, it remains a great challenge due to missing target annotations and the rich discrepancy between source and target distributions. We propose Spectral UDA (SUDA), an effective and efficient UDA technique that works in the spectral space and can generalize across different visual recognition tasks. SUDA addresses the UDA challenges from two perspectives. First, it introduces a spectrum transformer (ST) that mitigates inter-domain discrepancies by enhancing domain-invariant spectra while suppressing domain-variant spectra of source and target samples simultaneously. Second, it introduces multi-view spectral learning that learns useful unsupervised representations by maximizing mutual information among multiple ST-generated spectral views of each target sample. Extensive experiments show that SUDA achieves superior accuracy consistently across different visual tasks in object detection, semantic segmentation and image classification. Additionally, SUDA also works with the transformer-based network and achieves state-of-the-art performance on object detection.

The recently developed DEtection TRansformer (DETR) establishes a new object detection paradigm by eliminating a series of hand-crafted components. However, DETR suffers from extremely slow convergence, which increases the training cost significantly. We observe that the slow convergence is largely attributed to the complication in matching object queries with target features in different feature embedding spaces. This paper presents SAM-DETR, a Semantic-Aligned-Matching DETR that greatly accelerates DETR's convergence without sacrificing its accuracy. SAM-DETR addresses the convergence issue from two perspectives. First, it projects object queries into the same embedding space as encoded image features, where the matching can be accomplished efficiently with aligned semantics. Second, it explicitly searches salient points with the most discriminative features for semantic-aligned matching, which further speeds up the convergence and boosts detection accuracy as well. Being like a plug and play, SAM-DETR complements existing convergence solutions well yet only introduces slight computational overhead. Extensive experiments show that the proposed SAM-DETR achieves superior convergence as well as competitive detection accuracy. The implementation codes are publicly available at https://github.com/ZhangGongjie/SAM-DETR.

In a point cloud sequence, 3D object tracking aims to predict the location and orientation of an object in the current search point cloud given a template point cloud. Motivated by the success of transformers, we propose Point Tracking TRansformer (PTTR), which efficiently predicts high-quality 3D tracking results in a coarse-to-fine manner with the help of transformer operations. PTTR consists of three novel designs. 1) Instead of random sampling, we design Relation-Aware Sampling to preserve relevant points to given templates during subsampling. 2) Furthermore, we propose a Point Relation Transformer (PRT) consisting of a self-attention and a cross-attention module. The global self-attention operation captures long-range dependencies to enhance encoded point features for the search area and the template, respectively. Subsequently, we generate the coarse tracking results by matching the two sets of point features via cross-attention. 3) Based on the coarse tracking results, we employ a novel Prediction Refinement Module to obtain the final refined prediction. In addition, we create a large-scale point cloud single object tracking benchmark based on the Waymo Open Dataset. Extensive experiments show that PTTR achieves superior point cloud tracking in both accuracy and efficiency. Our code is available at https://github.com/Jasonkks/PTTR.

Dance choreography for a piece of music is a challenging task, having to be creative in presenting distinctive stylistic dance elements while taking into account the musical theme and rhythm. It has been tackled by different approaches such as similarity retrieval, sequence-to-sequence modeling and generative adversarial networks, but their generated dance sequences are often short of motion realism, diversity and music consistency. In this paper, we propose a Music-to-Dance with Optimal Transport Network (MDOT-Net) for learning to generate 3D dance choreographies from music. We introduce an optimal transport distance for evaluating the authenticity of the generated dance distribution and a Gromov-Wasserstein distance to measure the correspondence between the dance distribution and the input music. This gives a well defined and non-divergent training objective that mitigates the limitation of standard GAN training which is frequently plagued with instability and divergent generator loss issues. Extensive experiments demonstrate that our MDOT-Net can synthesize realistic and diverse dances which achieve an organic unity with the input music, reflecting the shared intentionality and matching the rhythmic articulation. Sample results are found at https://www.youtube.com/watch?v=dErfBkrlUO8.

The current LiDAR SLAM (Simultaneous Localization and Mapping) system suffers greatly from low accuracy and limited robustness when faced with complicated circumstances. From our experiments, we find that current LiDAR SLAM systems have limited performance when the noise level in the obtained point clouds is large. Therefore, in this work, we propose a general framework to tackle the problem of denoising and loop closure for LiDAR SLAM in complex environments with many noises and outliers caused by reflective materials. Current approaches for point clouds denoising are mainly designed for small-scale point clouds and can not be extended to large-scale point clouds scenes. In this work, we firstly proposed a lightweight network for large-scale point clouds denoising. Subsequently, we have also designed an efficient loop closure network for place recognition in global optimization to improve the localization accuracy of the whole system. Finally, we have demonstrated by extensive experiments and benchmark studies that our method can have a significant boost on the localization accuracy of the LiDAR SLAM system when faced with noisy point clouds, with a marginal increase in computational cost.

Leveraging StyleGAN's expressivity and its disentangled latent codes, existing methods can achieve realistic editing of different visual attributes such as age and gender of facial images. An intriguing yet challenging problem arises: Can generative models achieve counterfactual editing against their learnt priors? Due to the lack of counterfactual samples in natural datasets, we investigate this problem in a text-driven manner with Contrastive-Language-Image-Pretraining (CLIP), which can offer rich semantic knowledge even for various counterfactual concepts. Different from in-domain manipulation, counterfactual manipulation requires more comprehensive exploitation of semantic knowledge encapsulated in CLIP as well as more delicate handling of editing directions for avoiding being stuck in local minimum or undesired editing. To this end, we design a novel contrastive loss that exploits predefined CLIP-space directions to guide the editing toward desired directions from different perspectives. In addition, we design a simple yet effective scheme that explicitly maps CLIP embeddings (of target text) to the latent space and fuses them with latent codes for effective latent code optimization and accurate editing. Extensive experiments show that our design achieves accurate and realistic editing while driving by target texts with various counterfactual concepts.

Neural Radiance Fields (NeRF) has demonstrated very impressive performance in novel view synthesis via implicitly modelling 3D representations from multi-view 2D images. However, most existing studies train NeRF models with either reasonable camera pose initialization or manually-crafted camera pose distributions which are often unavailable or hard to acquire in various real-world data. We design VMRF, an innovative view matching NeRF that enables effective NeRF training without requiring prior knowledge in camera poses or camera pose distributions. VMRF introduces a view matching scheme, which exploits unbalanced optimal transport to produce a feature transport plan for mapping a rendered image with randomly initialized camera pose to the corresponding real image. With the feature transport plan as the guidance, a novel pose calibration technique is designed which rectifies the initially randomized camera poses by predicting relative pose transformations between the pair of rendered and real images. Extensive experiments over a number of synthetic and real datasets show that the proposed VMRF outperforms the state-of-the-art qualitatively and quantitatively by large margins.

This paper shows that masked generative adversarial network (MaskedGAN) is robust image generation learners with limited training data. The idea of MaskedGAN is simple: it randomly masks out certain image information for effective GAN training with limited data. We develop two masking strategies that work along orthogonal dimensions of training images, including a shifted spatial masking that masks the images in spatial dimensions with random shifts, and a balanced spectral masking that masks certain image spectral bands with self-adaptive probabilities. The two masking strategies complement each other which together encourage more challenging holistic learning from limited training data, ultimately suppressing trivial solutions and failures in GAN training. Albeit simple, extensive experiments show that MaskedGAN achieves superior performance consistently across different network architectures (e.g., CNNs including BigGAN and StyleGAN-v2 and Transformers including TransGAN and GANformer) and datasets (e.g., CIFAR-10, CIFAR-100, ImageNet, 100-shot, AFHQ, FFHQ and Cityscapes).

LiDAR point clouds, which are usually scanned by rotating LiDAR sensors continuously, capture precise geometry of the surrounding environment and are crucial to many autonomous detection and navigation tasks. Though many 3D deep architectures have been developed, efficient collection and annotation of large amounts of point clouds remain one major challenge in the analytics and understanding of point cloud data. This paper presents PolarMix, a point cloud augmentation technique that is simple and generic but can mitigate the data constraint effectively across various perception tasks and scenarios. PolarMix enriches point cloud distributions and preserves point cloud fidelity via two cross-scan augmentation strategies that cut, edit, and mix point clouds along the scanning direction. The first is scene-level swapping which exchanges point cloud sectors of two LiDAR scans that are cut along the LiDAR scanning direction. The second is instance-level rotation and paste which crops point instances from one LiDAR scan, rotates them by multiple angles (to create multiple copies), and paste the rotated point instances into other scans. Extensive experiments show that PolarMix achieves superior performance consistently across different perception tasks and scenarios. In addition, it can work as a plug-and-play for various 3D deep architectures and also performs well for unsupervised domain adaptation.

Recent progresses in domain adaptive semantic segmentation demonstrate the effectiveness of adversarial learning (AL) in unsupervised domain adaptation . However, most adversarial learning based methods align source and target distributions at a global image level but neglect the inconsistency around local image regions. This paper presents a novel multi-level adversarial network (MLAN) that aims to address inter-domain inconsistency at both global image level and local region level optimally. MLAN has two novel designs, namely, region-level adversarial learning (RL-AL) and co-regularized adversarial learning (CR-AL). Specifically, RL-AL models prototypical regional context-relations explicitly in the feature space of a labelled source domain and transfers them to an unlabelled target domain via adversarial learning. CR-AL fuses region-level AL and image-level AL optimally via mutual regularization . In addition, we design a multi-level consistency map that can guide domain adaptation in both input space (i.e., image-to-image translation) and output space (i.e., self-training) effectively. Extensive experiments show that MLAN outperforms the state-of-the-art with a large margin consistently across multiple datasets.

Various types of saliency cues exist, all of which can be instrumental in the foreground extraction. It brings us to an interesting problem of effectively combining them. Note that earlier works either fuse them in the spatial domain or introduce dedicated terms in the energy functions to cater to multiple cues. In contrast, this paper investigates the appearance domain and proposes a novel appearance fusion framework, which we refer to as AppFuse. It is an intuitive framework for fusing candidate appearance models into the desired one for an energy function. Thus, we do not require any alterations in the energy function anymore. Like any fusion strategy, the proposed framework also requires guidance, which we facilitate through reliability and mutual consensus phenomena. To demonstrate the efficacy, we leverage it to solve a foreground extraction problem named video co-localization, where we propose two novel concepts i) hierarchical co-saliency and ii) mask-specific proposals. Our fusion results ensure that similar objects get highlighted sufficiently to ensure localization simply by respecting our framework and different spatiotemporal constraints. Our exhaustive set of experiments using both hand-crafted and learned saliency cues reveal that our approach comfortably outperforms several competing localization methods on standard benchmark datasets.

Inferring the scene illumination from a single image is an essential yet challenging task in computer vision and computer graphics. Existing works estimate lighting by regressing representative illumination parameters or generating illumination maps directly. However, these methods often suffer from poor accuracy and generalization. This paper presents Geometric Mover's Light (GMLight), a lighting estimation framework that employs a regression network and a generative projector for effective illumination estimation. We parameterize illumination scenes in terms of the geometric light distribution, light intensity, ambient term, and auxiliary depth, which can be estimated by a regression network. Inspired by the earth mover's distance, we design a novel geometric mover's loss to guide the accurate regression of light distribution parameters. With the estimated light parameters, the generative projector synthesizes panoramic illumination maps with realistic appearance and high-frequency details. Extensive experiments show that GMLight achieves accurate illumination estimation and superior fidelity in relighting for 3D object insertion. The codes are available at https://github.com/fnzhan/Illumination-Estimation.

Unsupervised domain adaptive object detection aims to adapt detectors from a labelled source domain to an unlabelled target domain. Most existing works take a two-stage strategy that first generates region proposals and then detects objects of interest, where adversarial learning is widely adopted to mitigate the inter-domain discrepancy in both stages. However, adversarial learning may impair the alignment of well-aligned samples as it merely aligns the global distributions across domains. To address this issue, we design an uncertainty-aware domain adaptation network (UaDAN) that introduces conditional adversarial learning to align well-aligned and poorly-aligned samples separately in different manners. Specifically, we design an uncertainty metric that assesses the alignment of each sample and adjusts the strength of adversarial learning for well-aligned and poorly-aligned samples adaptively. In addition, we exploit the uncertainty metric to achieve curriculum learning that first performs easier image-level alignment and then more difficult instance-level alignment progressively. Extensive experiments over four challenging domain adaptive object detection datasets show that UaDAN achieves superior performance as compared with state-of-the-art methods.

Unsupervised domain adaptation (UDA) has achieved unprecedented success in improving the cross-domain robustness of object detection models. However, existing UDA methods largely ignore the instantaneous data distribution and the sampling strategy during model learning, which could deteriorate the feature representation given large domain shift. In this work, we propose a Self-Guided Adaptation (SGA) model, targeting at aligning feature representation and transferring object detection models across domains while considering the instantaneous alignment difficulty. The core of SGA is to calculate “hardness” factors for sample pairs indicating domain distance in a kernel space. With the hardness factor, the proposed SGA adaptively indicates the importance of samples and assigns them different constrains. Indicated by these hardness factors, Self-Guided Progressive Sampling (SPS) is implemented in an “easy-to-hard” way during model adaptation. Using multi-stage convolutional features, SGA is further aggregated to fully align hierarchical representations of detection models. Extensive experiments on commonly-used benchmarks show that SGA improves the state-of-the-art methods with significant margins especially on large domain shift cases.

Occluded person re-identification (re-ID) is a challenging task as different human parts may become invisible in cluttered scenes, making it hard to match person images of different identities. Most existing methods address this challenge by aligning spatial features of body parts according to semantic information (e.g. human poses) or feature similarities but this approach is complicated and sensitive to noises. This paper presents Matching on Sets (MoS), a novel method that positions occluded person re-ID as a set matching task without requiring spatial alignment. MoS encodes a person image by a pattern set as represented by a `global vector’ with each element capturing one specific visual pattern, and it introduces Jaccard distance as a metric to compute the distance between pattern sets and measure image similarity. To enable Jaccard distance over continuous real numbers, we employ minimization and maximization to approximate the operations of intersection and union, respectively. In addition, we design a Jaccard triplet loss that enhances the pattern discrimination and allows to embed set matching into deep neural networks for end-to-end training. In the inference stage, we introduce a conflict penalty mechanism that detects mutually exclusive patterns in the pattern union of image pairs and decreases their similarities accordingly. Extensive experiments over three widely used datasets (Market1501, DukeMTMC and Occluded-DukeMTMC) show that MoS achieves superior re-ID performance. Additionally, it is tolerant of occlusions and outperforms the state-of-the-art by large margins for Occluded-DukeMTMC.

Illumination estimation from a single image is critical in 3D rendering and it has been investigated extensively in the computer vision and computer graphic research community. On the other hand, existing works estimate illumination by either regressing light parameters or generating illumination maps that are often hard to optimize or tend to produce inaccurate predictions. We propose Earth Mover’s Light (EMLight), an illumination estimation framework that leverages a regression network and a neural projector for accurate illumination estimation. We decompose the illumination map into spherical light distribution, light intensity and the ambient term, and define the illumination estimation as a parameter regression task for the three illumination components. Motivated by the Earth Mover's distance, we design a novel spherical mover's loss that guides to regress light distribution parameters accurately by taking advantage of the subtleties of spherical distribution. Under the guidance of the predicted spherical distribution, light intensity and ambient term, the neural projector synthesizes panoramic illumination maps with realistic light frequency. Extensive experiments show that EMLight achieves accurate illumination estimation and the generated relighting in 3D object embedding exhibits superior plausibility and fidelity as compared with state-of-the-art methods.

Domain generalization aims to learn a generalizable model from a ‘known’ source domain for various ‘unknown’ target domains. It has been studied widely by domain randomization that transfers source images to different styles in spatial space for learning domain-agnostic features. However, most existing randomization methods use GANs that often lack of controls and even alter semantic structures of images undesirably. Inspired by the idea of JPEG that converts spatial images into multiple frequency components (FCs), we propose Frequency Space Domain Randomization (FSDR) that randomizes images in frequency space by keeping domain-invariant FCs (DIFs) and randomizing domain-variant FCs (DVFs) only. FSDR has two unique features: 1) it decomposes images into DIFs and DVFs which allows explicit access and manipulation of them and more controllable randomization; 2) it has minimal effects on semantic structures of images and domain-invariant features. We examined domain variance and invariance property of FCs statistically and designed a network that can identify and fuse DIFs and DVFs dynamically through iterative learning. Extensive experiments over multiple domain generalizable segmentation tasks show that FSDR achieves superior segmentation and its performance is even on par with domain adaptation methods that access target data in training.

Panoptic segmentation unifies semantic segmentation and instance segmentation which has been attracting increasing attention in recent years. However, most existing research was conducted under a supervised learning setup whereas unsupervised domain adaptive panoptic segmentation which is critical in different tasks and applications is largely neglected. We design a domain adaptive panoptic segmentation network that exploits inter-style consistency and inter-task regularization for optimal domain adaptive panoptic segmentation. The inter-style consistency leverages semantic invariance across the same image of the different styles which ‘ fabricates’ certain self-supervisions to guide the network to learn domain-invariant features. The inter-task regularization exploits the complementary nature of instance segmentation and semantic segmentation and uses it as a constraint for better feature alignment across domains. Extensive experiments over multiple domain adaptive panoptic segmentation tasks (e.g. synthetic-to-real and real-to-real) show that our proposed network achieves superior segmentation performance as compared with the state-of-the-art.

Despite the great success of GANs in images translation with different conditioned inputs such as semantic segmentation and edge maps, generating high-fidelity realistic images with reference styles remains a grand challenge in conditional image-to-image translation. This paper presents a general image translation framework that incorporates optimal transport for feature alignment between conditional inputs and style exemplars in image translation. The introduction of optimal transport mitigates the constraint of many-to-one feature matching significantly while building up accurate semantic correspondences between conditional inputs and exemplars. We design a novel unbalanced optimal transport to address the transport between features with deviational distributions which exists widely between conditional inputs and exemplars. In addition, we design a semantic-activation normalization scheme that injects style features of exemplars into the image translation process successfully. Extensive experiments over multiple image translation tasks show that our method achieves superior image translation qualitatively and quantitatively as compared with the state-of-the-art.

Unsupervised domain adaptation (UDA) involves a supervised loss in a labeled source domain and an unsupervised loss in an unlabeled target domain, which often faces more severe overfitting (than classical supervised learning) as the supervised source loss has clear domain gap and the unsupervised target loss is often noisy due to the lack of annotations. This paper presents RDA, a robust domain adaptation technique that introduces adversarial attacking to mitigate overfitting in UDA. We achieve robust domain adaptation by a novel Fourier adversarial attacking (FAA) method that allows large magnitude of perturbation noises but has minimal modification of image semantics, the former is critical to the effectiveness of its generated adversarial samples due to the existence of ‘domain gaps’. Specifically, FAA decomposes images into multiple frequency components (FCs) and generates adversarial samples by just perturbating certain FCs that capture little semantic information. With FAA-generated samples, the training can continue the ‘random walk’ and drift into an area with a flat loss landscape, leading to more robust domain adaptation. Extensive experiments over multiple domain adaptation tasks show that RDA can work with different computer vision tasks with superior performance.

Video semantic segmentation is an essential task for the analysis and understanding of videos. Recent efforts largely focus on supervised video segmentation by learning from fully annotated data, but the learnt models often experience clear performance drop while applied to videos of a different domain. This paper presents DA-VSN, a domain adaptive video segmentation network that addresses domain gaps in videos by temporal consistency regularization (TCR) for consecutive frames of target-domain videos. DA-VSN consists of two novel and complementary designs. The first is cross-domain TCR that guides the prediction of target frames to have similar temporal consistency as that of source frames (learnt from annotated source data) via adversarial learning. The second is intra-domain TCR that guides unconfident predictions of target frames to have similar temporal consistency as confident predictions of target frames. Extensive experiments demonstrate the superiority of our proposed domain adaptive video segmentation network which outperforms multiple baselines consistently by large margins.

Deep learning-based 3D object detection has achieved unprecedented success with the advent of large-scale autonomous driving datasets. However, drastic performance degradation remains a critical challenge for cross-domain deployment. In addition, existing 3D domain adaptive detection methods often assume prior access to the target domain annotations, which is rarely feasible in the real world. To address this challenge, we study a more realistic setting, unsupervised 3D domain adaptive detection, which only utilizes source domain annotations. 1) We first comprehensively investigate the major underlying factors of the domain gap in 3D detection. Our key insight is that geometric mismatch is the key factor of domain shift. 2) Then, we propose a novel and unified framework, Multi-Level Consistency Network (MLC-Net), which employs a teacher-student paradigm to generate adaptive and reliable pseudo-targets. MLC-Net exploits point-, instance- and neural statistics-level consistency to facilitate cross-domain transfer. Extensive experiments demonstrate that MLC-Net out-performs existing state-of-the-art methods (including those using additional target domain information) on standard benchmarks. Notably, our approach is detector-agnostic, which achieves consistent gains on both single- and two-stage 3D detectors. Code will be released.

Skeleton-based human action recognition has attracted increasing attention in recent years. However, most of the existing works focus on supervised learning which requiring a large number of annotated action sequences that are often expensive to collect. We investigate unsupervised representation learning for skeleton action recognition, and design a novel skeleton cloud colorization technique that is capable of learning skeleton representations from unlabeled skeleton sequence data. Specifically, we represent a skeleton action sequence as a 3D skeleton cloud and colorize each point in the cloud according to its temporal and spatial orders in the original (unannotated) skeleton sequence. Leveraging the colorized skeleton point cloud, we design an auto-encoder framework that can learn spatial-temporal features from the artificial color labels of skeleton joints effectively. We evaluate our skeleton cloud colorization approach with action classifiers trained under different configurations, including unsupervised, semi-supervised and fully-supervised settings. Extensive experiments on NTU RGB+D and NW-UCLA datasets show that the proposed method outperforms existing unsupervised and semi-supervised 3D action recognition methods by large margins, and it achieves competitive performance in supervised 3D action recognition as well.

Image inpainting aims to complete the missing or corrupted regions of images with realistic contents. The prevalent approaches adopt a hybrid objective of reconstruction and perceptual quality by using generative adversarial networks. However, the reconstruction loss and adversarial loss focus on synthesizing contents of different frequencies and simply applying them together often leads to inter-frequency conflicts and compromised inpainting. This paper presents WaveFill, a wavelet-based inpainting network that decomposes images into multiple frequency bands and fills the missing regions in each frequency band separately and explicitly. WaveFill decomposes images by using discrete wavelet transform (DWT) that preserves spatial information naturally. It applies L1 reconstruction loss to the decomposed low-frequency bands and adversarial loss to high-frequency bands, hence effectively mitigate inter-frequency conflicts while completing images in spatial domain. To address the inpainting inconsistency in different frequency bands and fuse features with distinct statistics, we design a novel normalization scheme that aligns and fuses the multi-frequency features effectively. Extensive experiments over multiple datasets show that WaveFill achieves superior image inpainting qualitatively and quantitatively.

Accurate lighting estimation is challenging yet critical to many computer vision and computer graphics tasks such as high-dynamic-range (HDR) relighting. Existing approaches model lighting in either frequency domain or spatial domain which is insufficient to represent the complex lighting conditions in scenes and tends to produce inaccurate estimation. This paper presents NeedleLight, a new lighting estimation model that represents illumination with needlets and allows lighting estimation in both frequency domain and spatial domain jointly. An optimal thresholding function is designed to achieve sparse needlets which trims redundant lighting parameters and demonstrates superior localization properties for illumination representation. In addition, a novel spherical transport loss is designed based on optimal transport theory which guides to regress lighting representation parameters with consideration of the spatial information. Furthermore, we propose a new metric that is concise yet effective by directly evaluating the estimated illumination maps rather than rendered images. Extensive experiments show that NeedleLight achieves superior lighting estimation consistently across multiple evaluation metrics as compared with state-of-the-art methods.

Music and dance have always co-existed as pillars of human activities, contributing immensely to the cultural, social, and entertainment functions in virtually all societies. Notwithstanding the gradual systematization of music and dance into two independent disciplines, their intimate connection is undeniable and one art-form often appears incomplete without the other. Recent research works have studied generative models for dance sequences conditioned on music. The dual task of composing music for given dances, however, has been largely overlooked. In this paper, we propose a novel extension, where we jointly model both tasks in a dual learning approach. To leverage the duality of the two modalities, we introduce an optimal transport objective to align feature embeddings, as well as a cycle consistency loss to foster overall consistency. Experimental results demonstrate that our dual learning framework improves individual task performance, delivering generated music compositions and dance choreographs that are realistic and faithful to the conditioned inputs.

Image inpainting is an underdetermined inverse problem, which naturally allows diverse contents to fill up the missing or corrupted regions realistically. Prevalent approaches using convolutional neural networks (CNNs) can synthesize visually pleasant contents, but CNNs suffer from limited perception fields for capturing global features. With image-level attention, transformers enable to model long-range dependencies and generate diverse contents with autoregressive modeling of pixel-sequence distributions. However, the unidirectional attention in autoregressive transformers is suboptimal as corrupted image regions may have arbitrary shapes with contexts from any direction. We propose BAT-Fill, an innovative image inpainting framework that introduces a novel bidirectional autoregressive transformer (BAT) for image inpainting. BAT utilizes the transformers to learn autoregressive distributions, which naturally allows the diverse generation of missing contents. In addition, it incorporates the masked language model like BERT, which enables bidirectionally modeling of contextual information of missing regions for better image completion. Extensive experiments over multiple datasets show that BAT-Fill achieves superior diversity and fidelity in image inpainting qualitatively and quantitatively.

Unsupervised domain adaptation aims to align a labeled source domain and an unlabeled target domain, but it requires to access the source data which often raises concerns in data privacy, data portability and data transmission efficiency. We study unsupervised model adaptation (UMA), or called Unsupervised Domain Adaptation without Source Data, an alternative setting that aims to adapt source-trained models towards target distributions without accessing source data. To this end, we design an innovative historical contrastive learning (HCL) technique that exploits historical source hypothesis to make up for the absence of source data in UMA. HCL addresses the UMA challenge from two perspectives. First, it introduces historical contrastive instance discrimination (HCID) that learns from target samples by contrasting their embeddings which are generated by the currently adapted model and the historical models. With the historical models, HCID encourages UMA to learn instance-discriminative target representations while preserving the source hypothesis. Second, it introduces historical contrastive category discrimination (HCCD) that pseudo-labels target samples to learn category-discriminative target representations. Specifically, HCCD re-weights pseudo labels according to their prediction consistency across the current and historical models. Extensive experiments show that HCL outperforms and state-of-the-art methods consistently across a variety of visual tasks and setups.

Automated defect inspection is critical for effective and efficient maintenance, repair, and operations in advanced manufacturing. On the other hand, automated defect inspection is often constrained by the lack of defect samples, especially when we adopt deep neural networks for this task. This paper presents Defect-GAN, an automated defect synthesis network that generates realistic and diverse defect samples for training accurate and robust defect inspection networks. Defect-GAN learns through defacement and restoration processes, where the defacement generates defects on normal surface images while the restoration removes defects to generate normal images. It employs a novel compositional layer-based architecture for generating realistic defects within various image backgrounds with different textures and appearances. It can also mimic the stochastic variations of defects and offer flexible control over the locations and categories of the generated defects within the image background. Extensive experiments show that Defect-GAN is capable of synthesizing various defects with superior diversity and fidelity. In addition, the synthesized defect samples demonstrate their effectiveness in training better defect inspection networks.

The human visual system can detect objects of unseen categories from merely a few examples. However, such capability remains absent in state-of-the-art detectors. To bridge this gap, several attempts have been proposed to perform few-shot detection by incorporating meta-learning techniques. Such methods can improve detection performance on unseen categories, but also add huge computational burden, and usually degrade detection performance on seen categories. In this paper, we present PNPDet, a novel Plug-and-Play Detector, for efficient few-shot detection without forgetting. It introduces a simple but effective architecture with separate sub-networks that disentangles the recognition of base and novel categories and prevents hurting performance on known categories while learning new concepts. Distance metric learning is further incorporated into sub-networks, consistently boosting detection performance for both base and novel categories. Experiments show that the proposed PNPDet can achieve comparable few-shot detection performance on unseen categories while not losing accuracy on seen categories, and also remain efficient and flexible at the same time.

We focus on unsupervised domain adaptation (UDA) in image segmentation . Existing works address this challenge largely by aligning inter-domain representations, which may lead over-alignment that impairs the semantic structures of images and further target-domain segmentation performance . We design a scale variance minimization (SVMin) method by enforcing the intra-image semantic structure consistency in the target domain. Specifically, SVMin leverages an intrinsic property that simple scale transformation has little effect on the semantic structures of images. It thus introduces certain supervision in the target domain by imposing a scale-invariance constraint while learning to segment an image and its scale-transformation concurrently. Additionally, SVMin is complementary to most existing UDA techniques and can be easily incorporated with consistent performance boost but little extra parameters. Extensive experiments show that our method achieves superior domain adaptive segmentation performance as compared with the state-of-the-art. Preliminary studies show that SVMin can be easily adapted for UDA-based image classification .

Single-image dehazing has been an important topic given the commonly occurred image degradation caused by adverse atmosphere aerosols. The key to haze removal relies on an accurate estimation of global air-light and the transmission map. Most existing methods estimate these two parameters using separate pipelines which reduces the efficiency and accumulates errors, thus leading to a suboptimal approximation, hurting the model interpretability, and degrading the performance. To address these issues, this article introduces a novel generative adversarial network (GAN) for single-image dehazing. The network consists of a novel compositional generator and a novel deeply supervised discriminator. The compositional generator is a densely connected network, which combines fine-scale and coarse-scale information. Benefiting from the new generator, our method can directly learn the physical parameters from data and recover clean images from hazy ones in an end-to-end manner. The proposed discriminator is deeply supervised, which enforces that the output of the generator to look similar to the clean images from low-level details to high-level structures. To the best of our knowledge, this is the first end-to-end generative adversarial model for image dehazing, which simultaneously outputs clean images, transmission maps, and air-lights. Extensive experiments show that our method remarkably outperforms the state-of-the-art methods. Furthermore, to facilitate future research, we create the HazeCOCO dataset which is currently the largest dataset for single-image dehazing.

Pose-based person image synthesis aims to generate a new image containing a person with a target pose conditioned on a source image containing a person with a specified pose. It is challenging as the target pose is arbitrary and often significantly differs from the specified source pose, which leads to large appearance discrepancy between the source and the target images. This paper presents the Pose Transform Generative Adversarial Network (PoT-GAN) for person image synthesis where the generator explicitly learns the transform between the two poses by manipulating the corresponding multi-scale feature maps. By incorporating the learned pose transform information into the multi-scale feature maps of the source image in a GAN architecture, our method reliably transfers the appearance of the person in the source image to the target pose with no need for any hard-coded spatial information depicting the change of pose. According to both qualitative and quantitative results, the proposed PoT-GAN demonstrates a state-of-the-art performance on three publicly available datasets for person image synthesis.

Benefiting from the powerful discriminative feature learning capability of convolutional neural networks (CNNs), deep learning techniques have achieved remarkable performance improvement for the task of salient object detection (SOD) in recent years. However, most existing deep SOD models do not fully exploit informative contextual features, which often leads to suboptimal detection performance in the presence of a cluttered background. This paper presents a context-aware attention module that detects salient objects by simultaneously constructing connections between each image pixel and its local and global contextual pixels. Specifically, each pixel and its neighbors bidirectionally exchange semantic information by computing their correlation coefficients, and this process aggregates contextual attention features both locally and globally. In addition, an attention-guided hierarchical network architecture is designed to capture fine-grained spatial details by transmitting contextual information from deeper to shallower network layers in a top-down manner. Extensive experiments on six public SOD datasets show that our proposed model demonstrates superior SOD performance against most of the current state-of-the-art models under different evaluation metrics.

Unsupervised domain adaptation (UDA) aims to mitigate the domain shift that occurs when transferring knowledge from a labeled source domain to an unlabeled target domain. While it has been studied for application in unsupervised person re-identification (ReID), the relations of feature distribution across the source and target domains remain underexplored, as they either ignore the local relations or omit the in-depth consideration of negative transfer when two domains do not share identical label spaces. In light of the above, this paper presents an innovative part-aware progressive adaptation network (PPAN) that exploits global and local relations for UDA-based ReID across domains. A multi-branch network is developed that explicitly learns discriminative feature representation from both whole-body images and body-part images under the supervision of a labeled source domain. Within each network branch, an independent UDA constraint is designed that aligns the global and local feature distributions from a labeled source domain with those of an unlabeled target domain. In addition, a novel progressive adaptation strategy (PAS) is designed that effectively alleviates the negative influence of outlier source identities. The proposed unsupervised ReID model is evaluated on five widely used datasets (Market-1501, DukeMTMC-reID, CUHK03, VIPeR and PRID), and experimental results demonstrate its superior robustness and effectiveness relative to state-of-the-art approaches.

Learning to adapt to a series of different goals in visual navigation is challenging. In this work, we present a model-embedded actor-critic architecture for the multigoal visual navigation task. To enhance the task cooperation in multigoal learning, we introduce two new designs to the reinforcement learning scheme: inverse dynamics model (InvDM) and multigoal colearning (MgCl). Specifically, InvDM is proposed to capture the navigation-relevant association between state and goal and provide additional training signals to relieve the sparse reward issue. MgCl aims at improving the sample efficiency and supports the agent to learn from unintentional positive experiences. Besides, to further improve the scene generalization capability of the agent, we present an enhanced navigation model that consists of two self-supervised auxiliary task modules. The first module, which is named path closed-loop detection, helps to understand whether the state has been experienced. The second one, namely the state-target matching module, tries to figure out the difference between state and goal. Extensive results on the interactive platform AI2-THOR demonstrate that the agent trained with the proposed method converges faster than state-of-the-art methods while owning good generalization capability. The video demonstration is available at https://vsislab.github.io/mgvn.

Annotating a qualitative large-scale facial expression dataset is extremely difficult due to the uncertainties caused by ambiguous facial expressions, low-quality facial images, and the subjectiveness of annotators. These uncertainties suspend the progress of large-scale Facial Expression Recognition (FER) in data-driven deep learning era. To address this problelm, this paper proposes to suppress the uncertainties by a simple yet efficient Self-Cure Network (SCN). Specifically, SCN suppresses the uncertainty from two different aspects: 1) a self-attention mechanism over FER dataset to weight each sample in training with a ranking regularization, and 2) a careful relabeling mechanism to modify the labels of these samples in the lowest-ranked group. Experiments on synthetic FER datasets and our collected WebEmotion dataset validate the effectiveness of our method. Results on public benchmarks demonstrate that our SCN outperforms current state-of-the-art methods with \textbf{88.14}\% on RAF-DB, \textbf{60.23}\% on AffectNet, and \textbf{89.35}\% on FERPlus.

Recent advances in Generative Adversarial Nets (GANs) have shown remarkable improvements for facial expression editing. However, current methods are still prone to generate artifacts and blurs around expression-intensive regions, and often introduce undesired overlapping artifacts while handling large-gap expression transformations such as transformation from furious to laughing. To address these limitations, we propose Cascade Expression Focal GAN (Cascade EF-GAN), a novel network that performs progressive facial expression editing with local expression focuses. The introduction of the local focus enables the Cascade EF-GAN to better preserve identity-related features and details around eyes, noses and mouths, which further helps reduce artifacts and blurs within the generated facial images. In addition, an innovative cascade transformation strategy is designed by dividing a large facial expression transformation into multiple small ones in cascade, which helps suppress overlapping artifacts and produce more realistic editing while dealing with large-gap expression transformations. Extensive experiments over two publicly available facial expression datasets show that our proposed Cascade EF-GAN achieves superior performance for facial expression editing.

Domain adaptive person re-identification (re-ID) is a challenging task, especially when person identities in target domains are unknown. Existing methods attempt to address this challenge by transferring image styles or aligning feature distributions across domains, whereas the rich unlabeled samples in target domains are not sufficiently exploited. This paper presents a novel augmented discriminative clustering (AD-Cluster) technique that estimates and augments person clusters in target domains and enforces the discrimination ability of re-ID models with the augmented clusters. AD-Cluster is trained by iterative density-based clustering, adaptive sample augmentation, and discriminative feature learning. It learns an image generator and a feature encoder which aim to maximize the intra-cluster diversity in the sample space and minimize the intra-cluster distance in the feature space in an adversarial min-max manner. Finally, AD-Cluster increases the diversity of sample clusters and improves the discrimination capability of re-ID models greatly. Extensive experiments over Market-1501 and DukeMTMC-reID show that AD-Cluster outperforms the state-of-the-art with large margins.

RGB-Infrared (IR) cross-modality person re-identification (re-ID), which aims to search an IR image in RGB gallery or vice versa, is a challenging task due to the large discrepancy between IR and RGB modalities. Existing methods address this challenge typically by aligning feature distributions or image styles across modalities, whereas the very useful similarities among gallery samples of the same modality (i.e. intra-modality sample similarities) are largely neglected. This paper presents a novel similarity inference metric (SIM) that exploits the intra-modality sample similarities to circumvent the cross-modality discrepancy targeting optimal cross-modality image matching. SIM works by successive similarity graph reasoning and mutual nearest-neighbor reasoning that mine cross-modality sample similarities by leveraging intra-modality sample similarities from two different perspectives. Extensive experiments over two cross-modality re-ID datasets (SYSU-MM01 and RegDB) show that SIM achieves significant accuracy improvement but with little extra training as compared with the state-of-the-art.

Anticipating the future motions of 3D articulate objects is challenging due to its non-linear and highly stochastic nature. Current approaches typically represent the skeleton of an articulate object as a set of 3D joints, which unfortunately ignores the relationship between joints, and fails to encode fine-grained anatomical constraints. Moreover, conventional recurrent neural networks, such as LSTM and GRU, are employed to model motion contexts, which inherently have difficulties in capturing long-term dependencies. To address these problems, we propose to explicitly encode anatomical constraints by modeling their skeletons with a Lie algebra representation. Importantly, a hierarchical recurrent network structure is developed to simultaneously encodes local contexts of individual frames and global contexts of the sequence. We proceed to explore the applications of our approach to several distinct quantities including human, fish, and mouse. Extensive experiments show that our approach achieves more natural and accurate predictions over state-of-the-art methods.

Automated recognition of texts in scenes has been a research challenge for years, largely due to the arbitrary text appearance variation in perspective distortion, text line curvature, text styles and different types of imaging artifacts. The recent deep networks are capable of learning robust representations with respect to imaging artifacts and text style changes, but still face various problems while dealing with scene texts with perspective and curvature distortions. This paper presents an end-to-end trainable scene text recognition system (ESIR) that iteratively removes perspective distortion and text line curvature as driven by better scene text recognition performance. An innovative rectification network is developed, where a line-fitting transformation is designed to estimate the pose of text lines in scenes. Additionally, an iterative rectification framework is developed which corrects scene text distortions iteratively towards a fronto-parallel view. The ESIR is also robust to parameter initialization and easy to train, where the training needs only scene text images and word-level annotations as required by most scene text recognition systems. Extensive experiments over a number of public datasets show that the proposed ESIR is capable of rectifying scene text distortions accurately, achieving superior recognition performance for both normal scene text images and those suffering from perspective and curvature distortions.

Recent advances in generative adversarial networks (GANs) have shown great potentials in realistic image synthesis whereas most existing works address synthesis realism in either appearance space or geometry space but few in both. This paper presents an innovative Spatial Fusion GAN (SF-GAN) that combines a geometry synthesizer and an appearance synthesizer to achieve synthesis realism in both geometry and appearance spaces. The geometry synthesizer learns contextual geometries of background images and transforms and places foreground objects into the background images unanimously. The appearance synthesizer adjust the color, brightness and styles of the foreground objects and embeds them into background images harmoniously, where a guided filter is incorporated for detail preserving. The two synthesizers are inter-connected as mutual references which can be trained end-to-end with little supervision. The SF-GAN has been evaluated in two tasks: (1) realistic scene text image synthesis for training better recognition models; (2) glass and hat wearing for realistic matching glasses and hats with real portraits. Qualitative and quantitative comparisons with the state-of-the-art demonstrate the superiority of the proposed SF-GAN.

Recent adversarial learning research has achieved very impressive progress for modelling cross-domain data shifts in appearance space but its counterpart in modelling cross-domain shifts in geometry space lags far behind. This paper presents an innovative Geometry-Aware Domain Adaptation Network (GA-DAN) that is capable of modelling cross-domain shifts concurrently in both geometry space and appearance space and realistically converting images across domains with very different characteristics. In the proposed GA-DAN, a novel multi-modal spatial learning structure is designed which can convert a source-domain image into multiple images of different spatial views as in the target domain. A new disentangled cycle-consistency loss is introduced which balances the cycle consistency and greatly improves the concurrent learning in both appearance and geometry spaces. The proposed GA-DAN has been evaluated for the classic scene text detection and recognition tasks, and experiments show that the domain-adapted images achieve superior scene text detection and recognition performance while applied to network training.

The ICDAR 2019 Challenge on "Scanned receipts OCR and key information extraction" (SROIE) covers important aspects related to the automated analysis of scanned receipts. The SROIE tasks play a key role in many document analysis systems and hold significant commercial potential. Although a lot of work has been published over the years on administrative document analysis, the community has advanced relatively slowly, as most datasets have been kept private. One of the key contributions of SROIE to the document analysis community is to offer a first, standardized dataset of 1000 whole scanned receipt images and annotations, as well as an evaluation procedure for such tasks. The Challenge is structured around three tasks, namely Scanned Receipt Text Localization (Task 1), Scanned Receipt OCR (Task 2) and Key Information Extraction from Scanned Receipts (Task 3). The competition opened on 10th February, 2019 and closed on 5th May, 2019. We received 29, 24 and 18 valid submissions received for the three competition tasks, respectively. This report presents the competition datasets, define the tasks and the evaluation protocols, offer detailed submission statistics, as well as an analysis of the submitted performance. While the tasks of text localization and recognition seem to be relatively easy to tackle, it is interesting to observe the variety of ideas and approaches proposed for the information extraction task. According to the submissions' performance we believe there is still margin for improving information extraction performance, although the current dataset would have to grow substantially in following editions. Given the success of the SROIE competition evidenced by the wide interest generated and the healthy number of submissions from academic, research institutes and industry over different countries, we consider that the SROIE competition can evolve into a useful resource for the community, drawing further attention and promoting research and development efforts in this field.

Chinese scene text reading is one of the most challenging problems in computer vision and has attracted great interest. Different from English text, Chinese has more than 6000 commonly used characters and Chinese characters can be arranged in various layouts with numerous fonts. The Chinese signboards in street view are a good choice for Chinese scene text images since they have different backgrounds, fonts and layouts. We organized a competition called ICDAR2019-ReCTS, which mainly focuses on reading Chinese text on signboard. This report presents the final results of the competition. A large-scale dataset of 25,000 annotated signboard images, in which all the text lines and characters are annotated with locations and transcriptions, were released. Four tasks, namely character recognition, text line recognition, text line detection and end-to-end recognition were set up. Besides, considering the Chinese text ambiguity issue, we proposed a multi ground truth (multi-GT) evaluation method to make evaluation fairer. The competition started on March 1, 2019 and ended on April 30, 2019. 262 submissions from 46 teams are received. Most of the participants come from universities, research institutes, and tech companies in China. There are also some participants from the United States, Australia, Singapore, and Korea. 21 teams submit results for Task 1, 23 teams submit results for Task 2, 24 teams submit results for Task 3, and 13 teams submit results for Task 4. The official website for the competition is http://rrc.cvc.uab.es/?ch=12.

Learning to adapt to a series of different goals in visual navigation is challenging. In this work, we present a model-embedded actor-critic architecture for the multi-goal visual navigation task. To enhance the task cooperation in multi-goal learning, we introduce two new designs to the reinforcement learning scheme: inverse dynamics model (InvDM) and multi-goal co-learning (MgCl). Specifically, InvDM is proposed to capture the navigation-relevant association between state and goal, and provide additional training signals to relieve the sparse reward issue. MgCl aims at improving the sample efficiency and supports the agent to learn from unintentional positive experiences. Extensive results on the interactive platform AI2-THOR demonstrate that the proposed method converges faster than state-of-the-art methods while producing more direct routes to navigate to the goal. The video demonstration is available at: https://youtube.com/channel/UCtpTMOsctt3yPzXqe_JMD3w/videos.

State-of-the-art scene text detection techniques predict quadrilateral boxes that are prone to localization errors while dealing with straight or curved text lines of different orientations and lengths in scenes. This paper presents a novel multi-scale shape regression network (MSR) that is capable of locating text lines of different lengths, shapes and curvatures in scenes. The proposed MSR detects scene texts by predicting dense text boundary points that inherently capture the location and shape of text lines accurately and are also more tolerant to the variation of text line length as compared with the state of the arts using proposals or segmentation. Additionally, the multi-scale network extracts and fuses features at different scales which demonstrates superb tolerance to the text scale variation. Extensive experiments over several public datasets show that the proposed MSR obtains superior detection performance for both curved and straight text lines of different lengths and orientations.

Accurate and robust detection of multi-class objects in optical remote sensing images is essential to many real-world applications, such as urban planning, traffic control, searching, and rescuing. However, the state-of-the-art object detection techniques designed for images captured using ground-level sensors usually experience a sharp performance drop when directly applied to remote sensing images, largely due to the object appearance differences in remote sensing images in terms of sparse texture, low contrast, arbitrary orientations, and large-scale variations. This paper presents a novel object detection network [(context-aware detection network (CAD-Net)] that exploits attention-modulated features as well as global and local contexts to address the new challenges in detecting objects from remote sensing images. The proposed CAD-Net learns global and local contexts of objects by capturing their correlations with the global scene (at scene level) and the local neighboring objects or features (at object level), respectively. In addition, it designs a spatial-and-scale-aware attention module that guides the network to focus on more informative regions and features as well as more appropriate feature scales. Experiments over two publicly available object detection data sets for remote sensing images demonstrate that the proposed CAD-Net achieves superior detection performance. The implementation codes will be made publicly available for facilitating future works.

The availability of large-scale annotated data and uneven separability of different data categories become two major impediments of deep learning for image classification. In this paper, we present a Semi-Supervised Hierarchical Convolutional Neural Network (SS-HCNN) to address these two challenges. A large-scale unsupervised maximum margin clustering technique is designed, which splits images into a number of hierarchical clusters iteratively to learn cluster-level CNNs at parent nodes and category-level CNNs at leaf nodes. The splitting uses the similarity of CNN features to group visually similar images into the same cluster, which relieves the uneven data separability constraint. With the hierarchical cluster-level CNNs capturing certain high-level image category information, the category-level CNNs can be trained with a small amount of labelled images, and this relieves the data annotation constraint. A novel cluster splitting criterion is also designed which automatically terminates the image clustering in the tree hierarchy. The proposed SS-HCNN has been evaluated on the CIFAR-100 and ImageNet classification datasets. Experiments show that the SS-HCNN trained using a portion of labelled training images can achieve comparable performance with other fully trained CNNs using all labelled images. Additionally, the SS-HCNN trained using all labelled images clearly outperforms other fully trained CNNs.

The marriage between the deep convolutional neural network (CNN) and region proposals has made breakthroughs for object detection in recent years. While the discriminative object features are learned via a deep CNN for classification, the large intra-class variation and deformation still limit the performance of the CNN based object detection. We propose a subcategory-aware CNN (S-CNN) to solve the object intra-class variation problem. In the proposed technique, the training samples are first grouped into multiple subcategories automatically through a novel instance sharing maximum margin clustering process. A multi-component Aggregated Channel Feature (ACF) detector is then trained to produce more latent training samples, where each ACF component corresponds to one clustered subcategory. The produced latent samples together with their subcategory labels are further fed into a CNN classifier to filter out false proposals for object detection. An iterative learning algorithm is designed for the joint optimization of image subcategorization, multi-component ACF detector, and subcategory-aware CNN classifier. Experiments on INRIA Person dataset, Pascal VOC 2007 dataset and MS COCO dataset show that the proposed technique clearly outperforms the state-of-the-art methods for generic object detection.

This paper presents a new technique for hyperspectral image (HSI) classification by using superpixel guided deep-sparse-representation learning. The proposed technique constructs a hierarchical architecture by exploiting the sparse coding to learn the HSI representation. Specifically, a multiple-layer architecture using different superpixel maps is designed, where each superpixel map is generated by downsampling the superpixels gradually along with enlarged spatial regions for labeled samples. In each layer, sparse representation of pixels within every spatial region is computed to construct a histogram via the sum-pooling with l1normalization. Finally, the representations (features) learned from the multiple-layer network are aggregated and trained by a support vector machine classifier. The proposed technique has been evaluated over three public HSI data sets, including the Indian Pines image set, the Salinas image set, and the University of Pavia image set. Experiments show superior performance compared with the state-of-the-art methods.

The requiring of large amounts of annotated training data has become a common constraint on various deep learning systems. In this paper, we propose a weakly supervised scene text detection method (WeText) that trains robust and accurate scene text detection models by learning from unannotated or weakly annotated data. With a "light" supervised model trained on a small fully annotated dataset, we explore semi-supervised and weakly supervised learning on a large unannotated dataset and a large weakly annotated dataset, respectively. For the unsupervised learning, the light supervised model is applied to the unannotated dataset to search for more character training samples, which are further combined with the small annotated dataset to retrain a superior character detection model. For the weakly supervised learning, the character searching is guided by high-level annotations of words/text lines that are widely available and also much easier to prepare. In addition, we design an unified scene character detector by adapting regression based deep networks, which greatly relieves the error accumulation issue that widely exists in most traditional approaches. Extensive experiments across different unannotated and weakly annotated datasets show that the scene text detection performance can be clearly boosted under both scenarios, where the weakly supervised learning can achieve the state-of-the-art performance by using only 229 fully annotated scene text images.

Given a video clip, action proposal aims to quickly generate a number of spatio-temporal tubes that enclose candidate human activities. Recently, the regression-based networks and long-term recurrent convolutional network (L-RCN) have demonstrated superior performance in object detection and action recognition. However, the regression-based detectors perform inference without considering the temporal context among neighboring frames, and the LRC-N using global visual percepts lacks the capability to capture local temporal dynamics. In this paper, we present a novel framework called TORNADO for human action proposal detection in un-trimmed video clips. Specifically, we propose a spatio-temporal convolutional network that combines the advantages of regression-based detector and L-RCN by empowering Convolutional LSTM with regression capability. Our approach consists of a temporal convolutional regression network (T-CRN) and a spatial regression network (S-CRN) which are trained end-to-end on both RGB and optical flow streams. They fuse appearance, motion and temporal contexts to regress the bounding boxes of candidate human actions simultaneously in 28 FPS. The action proposals are constructed by solving dynamic programming with peak trimming of the generated action boxes. Extensive experiments on the challenging UCF-101 and UCF-Sports datasets show that our method achieves superior performance as compared with the state-of-the-arts.

Chinese is the most widely used language in the world. Algorithms that read Chinese text in natural images facilitate applications of various kinds. Despite the large potential value, datasets and competitions in the past primarily focus on English, which bares very different characteristics than Chinese. This report introduces RCTW, a new competition that focuses on Chinese text reading. The competition features a large-scale dataset with over 12,000 annotated images. Two tasks, namely text localization and end-to-end recognition, are set up. The competition took place from January 20 to May 31, 2017. 23 valid submissions were received from 19 teams. This report includes dataset description, task definitions, evaluation protocols, and results summaries and analysis. Through this competition, we call for more future research on the Chinese text reading problem.

Automatic reading texts in scenes is an attracting increasing interest in recent years due to various context awareness applications. Leverage on the advantages of object proposal in generic object detection, we propose a max-pooling based scene text proposal technique aiming for automatic extraction of texts in scenes. Given a scene image, a max-pooling based grouping technique is designed to search for scene text proposals within a feature map which is computed from image edges. Searched proposals are then ranked by a scoring function that is defined based on the histogram of oriented gradient. The proposed technique has been evaluated on two publicly available scene text datasets, including the ICDAR2015 dataset and the Street View Text (SVT) dataset. Experiments show that the proposed technique obtains superior proposal performance as compared with state-of-the-arts, especially when a small number of proposals is selected. In addition, it also obtains state-of-the-art scene text spotting when integrated with a scene text recognition model.

In recent years, text recognition has achieved remarkable success in recognizing scanned document text. However, word recognition in natural images is still an open problem, which generally requires time consuming post-processing steps. We present a novel architecture for individual word detection in scene images based on semantic segmentation. Our contributions are twofold: the concept of WordFence, which detects border areas surrounding each individual word and a novel pixelwise weighted softmax loss function which penalizes background and emphasizes small text regions. WordFence ensures that each word is detected individually, and the new loss function provides a strong training signal to both text and word border localization. The proposed technique avoids intensive post-processing, producing an end-to-end word detection system. We achieve superior localization recall on common benchmark datasets - 92% recall on ICDAR11 and ICDAR13 and 63% recall on SVT. Furthermore, our end-to-end word recognition system achieves state-of-the-art 86% F-Score on ICDAR13.

Change detection for aerial and remote sensing imageries is an important research topic with a wide range of applications in urban and environmental studies, emergency management, etc. It is a challenging problem due to various types of acquisition or environmental noises in the captured images. In this paper, we propose a saliency-based change detection technique that makes use of two-dimensional within-images and between-images co-occurrence histogram saliency. Our experimental results show that the proposed method can effectively detect true changes of semantic interest while suppressing false changes due to acquisition or environmental noises.

In this paper, we propose a new approach for searching action proposals in unconstrained videos. Our method first produces snippet action proposals by combining state-of-the-art YOLO detector (Static YOLO) and our regression based RNN detector (Recurrent YOLO). Then, these short action proposals are integrated to form final action proposals by solving two-pass dynamic programming which maximizes actioness score and temporal smoothness concurrently. Our experimental comparison with other state-of-the-arts on challenging UCF101 dataset shows that our method advances state-of-the-art proposal generation performance while maintaining low computational cost.

Visual attention modelling is an important research topic with a wide range of applications in visual tracking, perceptual quality assessment, re-targeting, video summarization, etc. In this paper, we propose a visual attention model that captures both bottom-up spatial-temporal saliency and top-down objectness. Leveraging on co-occurrence histograms, the proposed model captures a number of low-level cues including contrast, gradient, as well as, magnitude and gradient of optical flow. Additionally, the proposed model incorporates mid-level objectness cue which helps to boost the modelling performance greatly. The proposed model obtained superior AUC-ROCs when evaluated over the ASCMN dataset and the UCF Sports Action dataset.

We design an Enriched Deep Recurrent Visual Attention Model (EDRAM) — an improved attention-based architecture for multiple object recognition. The proposed model is a fully differentiable unit that can be optimized end-to-end by using Stochastic Gradient Descent (SGD). The Spatial Transformer (ST) was employed as visual attention mechanism which allows to learn the geometric transformation of objects within images. With the combination of the Spatial Transformer and the powerful recurrent architecture, the proposed EDRAM can localize and recognize objects simultaneously. EDRAM has been evaluated on two publicly available datasets including MNIST Cluttered (with 70K cluttered digits) and SVHN (with up to 250k real world images of house numbers). Experiments show that it obtains superior performance as compared with the state-of-the-art models.

Text proposal has been gaining interest in recent years due to the great success of object proposal in categoriesindependent object localization. In this paper, we present a novel text-specific proposal technique that provides superior bounding boxes for accurate text localization in scenes. The proposed technique, which we call Text Edge Box (TEB), uses a binary edge map, a gradient map and an orientation map of an image as inputs. Connected components are first found within the binary edge map, which are scored by two proposed low-cue text features that are extracted in the gradient map and the orientation map, respectively. These scores present text probability of connected components and are aggregated in a text edge image. Scene texts proposals are finally generated by grouping the connected components and estimating their likelihood of being words. The proposed TEB has been evaluated on the two public scene text datasets: the Robust Reading Competition 2013 dataset (ICDAR 2013) dataset and the Street View Text (SVT) dataset. Experiments show that the proposed TEB outperforms the state-of-the-art techniques greatly.

Vehicle detection and vehicle viewpoint estimation are both crucial for assistive and autonomous driving systems. In this paper, we propose a soft discriminative mixture of viewpoint (SDMoV) models for joint vehicle detection and vehicle viewpoint estimation. The proposed SDMoV model is learned in two steps. First, a discriminative viewpoint-specific component model, which aims to maximize vehicle viewpoint classification accuracy, is learned for each cluster of vehicle images with similar viewpoint. Second, a new soft margin objective function, which aims to maximize vehicle detection accuracy, is designed to retrain these component models into a mixture of viewpoint models. The proposed SDMoV model is capable of detecting vehicles and estimating their viewpoints simultaneously. Experiments on three state-of-the-art datasets show that the proposed SDMoV model achieves superior accuracy for both vehicle detection and vehicle viewpoint estimation tasks.

It is important for a moving observer to be able to identify his/her surrounding objects and determine whether these objects are moving or stationary, which is called object-level motion detection. Detecting object-level motion from moving cameras is a difficult problem to solve for collision-free navigation due to the dual motion introduced by the mixture of the camera motion and the object motion. This paper presents a novel technique that detects object-level motion from a freely moving camera using only two consecutive video frames. A context-aware motion descriptor (CMD) is designed based on the object’s moving speed and moving direction relative to that of the moving camera. The CMD employs the contextual information, e.g., the optical flow of the image background surrounding the moving object of interest, which describes the object motion behavior better than other contexts such as the camera’s GPS and direction. The inconsistency between the histogram of oriented optical flow of the object and its surrounding background is measured for the object-level motion detection. The proposed technique has been evaluated over two types of widely studied objects, i.e., vehicles and humans that are captured with different sizes, moving speeds, and image backgrounds using a moving camera. Experiments on challenging real-world videos show promising performance in object-level motion detection.

Oil spill inspection is critical to the marine and coastal ecosystems, and has been widely studied by various remote sensing technologies, such as synthetic aperture radar and hyperspectral. To improve the temporal resolution and the inspection flexibility, we propose a novel aerial image-based system that can find oil spills timely from images captured using onboard optical cameras installed in unmanned aerial vehicle or airplanes. In particular, a subcategory-aware feature selection (FS) and support vector machine (SVM) joint optimization technique is proposed to learn a discriminative model that can tell the existence of oil spills within an optical image of the marine surface. A set of color-based features is first extracted and concatenated together to characterize the oil spill incidence in an image, where a new color autocorrelogram is designed, which can better describe each color's spatial distribution in an image. Furthermore, subcategory-aware joint FS and SVM optimization technique is designed, which is capable of generating the optimal feature subset and SVM component models. Experiments on a set of real-world marine surface images show that the proposed technique outperforms the state-of-the-art techniques and achieves promising results for aerial image-based oil spill inspection.

This paper proposes a novel unsupervised feature learning algorithm for land-use scene recognition on very high resolution remote sensing imagery. The proposed technique utilizes a multipath sparse coding architecture in order to capture multiple aspects of discriminative structures within complex remote sensing sceneries. Unlike the previous sparse coding and bag-of-visual-words-based techniques that rely on the handcrafted feature descriptors such as scale-invariant feature transform, the proposed technique extracts dense low-level features from the raw data, including the visual (RGB) data and near-infrared (NIR) data, using image patches of varying sizes at different layers. The proposed technique has been evaluated on three data sets, including the 21-category UC Merced landuse RGB data set with a 1-ft spatial resolution, the 9-category ground scene RGB-NIR data set, and the 10-category Singapore land-use RGB-NIR data set with a 0.5-m spatial resolution. The experimental results show that the proposed technique outperforms the state-of-the-art methods.

RGBD scene recognition has attracted increasingly attention due to the rapid development of depth sensors and their wide application scenarios. While many research has been conducted, most work used hand-crafted features which are difficult to capture high-level semantic structures. Recently, the feature extracted from deep convolutional neural network has produced state-of-the-art results for various computer vision tasks, which inspire researchers to explore incorporating CNN learned features for RGBD scene understanding. On the other hand, most existing work combines rgb and depth features without adequately exploiting the consistency and complementary information between them. Inspired by some recent work on RGBD object recognition using multi-modal feature fusion, we introduce a novel discriminative multi-modal fusion framework for rgbd scene recognition for the first time which simultaneously considers the inter-and intra-modality correlation for all samples and meanwhile regularizing the learned features to be discriminative and compact. The results from the multimodal layer can be back-propagated to the lower CNN layers, hence the parameters of the CNN layers and multimodal layers are updated iteratively until convergence. Experiments on the recently proposed large scale SUN RGB-D datasets show that our method achieved the state-of-the-art without any image segmentation.

Localizing, identifying, and extracting humans with consistent appearance jointly from a personal photo stream is an important problem and has wide applications. The strong variations in foreground and background and irregularly occurring foreground humans make this realistic problem challenging. Inspired by advancements in object detection, scene understanding, and image cosegmentation, we explore explicit constraints to label and segment human objects rather than other nonhuman objects and “stuff.” We refer to such a problem as multiple human identification and cosegmentation (MHIC). To identify specific human subjects, we propose an efficient human instance detector by combining an extended color line model with a poselet-based human detector. Moreover, to capture high-level human shape information, a novel soft shape cue is proposed. It is initialized by the human detector, then further enhanced through a generalized geodesic distance transform, and finally refined with a joint bilateral filter. We also propose to capture the rich feature context around each pixel by using an adaptive cross-region data structure, which gives a higher discriminative power than a single pixel-based estimation. The high-level object cues from the detector and the shape are then integrated with the low-level pixel cues and midlevel contour cues into a principled conditional random field (CRF) framework, which can be efficiently solved by using fast graph cut algorithms. We evaluate our method over a newly created NTU-MHIC human dataset, which contains 351 images with manually annotated groundtruth segmentation. Both visual and quantitative results demonstrate that our method achieves state-of-the-art performance for the MHIC task.

Recent top performing methods in PASCAL VOC [6] and ImageNet [13] make use of object proposal to replace exhaustive window search. Object proposal’s effectiveness is rooted in the assumption that there are general cues to differentiate objects from the background. Since the very first work by Alexe et al. [1], many object proposal methods have been proposed [2, 3, 4, 5, 7, 8, 10, 11, 12, 14, 15] and tested on various large scale datasets [6, 9, 13], and their overall detection rates versus different thresholds or window number have also been reported. Yet such partial performance summaries give us little idea of a method’s strengths and weaknesses for further improvement, and users are still facing difficulties in choosing methods for their applications. Therefore, more detailed analysis of existing state-of-the-arts is critical for future research and applications. Our contributions can be summarized in three aspects. First, we investigate the influence of object-level characteristics over state-of-the-art object proposal methods for the first time. Although there are some similar works in categorical object detection, few research has been conducted on object proposal side to the best of our knowledge. Second, we introduce the concept of localization latency to evaluate a method’s localization efficiency and accuracy. Third, we create a fully annotated PASCAL VOC dataset with various object-level characteristics to facilitate our analysis. The annotations take us nearly one month’s time which will be released to facilitate further related research. Our experiments are based on PASCAL VOC2007 test set, which has been widely used in evaluating object proposal methods. A proposed window B is treated as detected if its Intersection-over-Union (IoU) with a ground truth bounding box B: IoU(B,B) = area(B ∩ B) area(B ∪ B) is above a certain threshold T . We first study the localization accuracy of the existing methods. The region based methods have higher localization accuracy than window based methods. MCG and SelectiveSearch are the top performing region based methods, though window based EdgeBox shows comparable performance. The localization accuracy for region based methods are similar. One potential explanation is that all region based methods follow similar pipeline by grouping superpixels with either learned or handcrafted edge measures. A good object proposal method should not only produce candidates with high accuracy, but also use as less windows as possible. To summarize a method’s performance in terms of the accuracy and window number, we propose the localization latency metric:

The prevalent scene text detection approach follows four sequential steps comprising character candidate detection, false character candidate removal, text line extraction, and text line verification. However, errors occur and accumulate throughout each of these sequential steps which often lead to low detection performance. To address these issues, we propose a unified scene text detection system, namely Text Flow, by utilizing the minimum cost (min-cost) flow network model. With character candidates detected by cascade boosting, the min-cost flow network model integrates the last three sequential steps into a single process which solves the error accumulation problem at both character level and text line level effectively. The proposed technique has been tested on three public datasets, i.e, ICDAR2011 dataset, ICDAR2013 dataset and a multilingual dataset and it outperforms the state-of-the-art methods on all three datasets with much higher recall and F-score. The good performance on the multilingual dataset shows that the proposed technique can be used for the detection of texts in different languages.

Results of the ICDAR 2015 Robust Reading Competition are presented. A new Challenge 4 on Incidental Scene Text has been added to the Challenges on Born-Digital Images, Focused Scene Images and Video Text. Challenge 4 is run on a newly acquired dataset of 1,670 images evaluating Text Localisation, Word Recognition and End-to-End pipelines. In addition, the dataset for Challenge 3 on Video Text has been substantially updated with more video sequences and more accurate ground truth data. Finally, tasks assessing End-to-End system performance have been introduced to all Challenges. The competition took place in the first quarter of 2015, and received a total of 44 submissions. Only the tasks newly introduced in 2015 are reported on. The datasets, the ground truth specification and the evaluation protocols are presented together with the results and a brief summary of the participating methods.

Recognition of handwritten text is a useful technique that can be applied in different applications, such as signature recognition, bank check recognition, etc. However, the off-line handwritten text recognition in an unconstrained situation is still a very challenging task due to the high complexity of text strokes and image background. This paper presents a novel segmented handwritten text recognition technique that ensembles recurrent neural network (RNN) classifiers. Two RNN models are first trained that take advantage of the widely used geometrical feature and the Histogram of Oriented Gradient (HOG) feature, respectively. Given a handwritten word image, the optimal recognition result is then obtained by integrating the two trained RNN models together with a lexicon. Experiments on public datasets show the superior performance of our proposed technique.

Text segmentation provides important clues for the accurate identification of character locations and the analysis of character properties such as shape estimation and texture synthesis. In this paper, we propose a robust text segmentation method that employs Markov Random Field (MRF) and use graph cut algorithms to solve the energy minimization problem. To effectively select accurate seeds to boost the text segmentation performance, stroke feature transform is adopted to robustly identify text seeds and text edges. Background seeds are obtained near the text edges in order to well preserve the text boundaries. The energy functions are defined as an MRF consisting of data energy and smoothness energy which can be efficiently solved by graph cut algorithms. One distinctive property of the proposed technique is that it can identify more distinctive seeds so that only one cut is needed to well separate the text regions from the background, hence much faster than the existing iterative graph cut approach. Experiments on ICDAR 2003 and ICDAR 2011 datasets show that the proposed technique obtains superior performance on both pixel level and atom level segmentation.

Vehicle viewpoint estimation plays an important role for intelligent transportation systems. We present an effective vehicle viewpoint estimation technique by utilizing the spatial location information of root and part objects detected in vehicle images via deformable part-based model (DPM). The viewpoint-aware spatial distribution of each part relative to the root is learned using the Gaussian mixture model. The discriminative capability of each part for each viewpoint is then estimated through measuring the Kullback Leibler divergence between pairwise viewpoint-aware spatial distributions. The discriminative information is finally used to compute the likelihood that the detected vehicle belongs to each viewpoint. Experimental results on a benchmark dataset demonstrate the superior performance of the proposed vehicle viewpoint estimation technique.

Automatic lane marking detection plays an important role in intelligent transportation systems. We present an effective lane marking detection technique that utilizes the context-aware information of lane marking on the urban roads. The proposed technique consists of two innovations. First, the context-aware color, texture and shape features which characterise both lane markings and their road context are designed to represent the lane markings on the road surface. Second, a hard negative mining technique is developed based on the Maximum Stable Extreme Region (MSER) detector and adaboost training. Experiments on a real world dataset demonstrate the superior performance of the proposed approach.

Automatic vegetation coverage detection plays a key role for monitoring and management of land usage, environmental variation, and urban planning. This paper presents a novel vegetation coverage detection technique for very high resolution multi-spectral satellite imagery. The proposed technique consists of two stages including a supervised patch-level scoring stage and an unsupervised pixel-level classification stage. In the first stage, a support vector regression (SVR) technique is developed which scores each image patch and generates a coarse patch-level vegetation map. In the second stage, an unsupervised pixel-level vegetation classification technique is developed, which produces a more detailed vegetation map by re-scoring those uncertain pixels based on the computed SVR scores. Experiments on very high resolution multi-spectral satellite images show that the proposed technique outperforms the state-of-the-art methods in both patch-level and pixel-level vegetation detection.

This paper presents a codebook learning based mobile landmark recognition technique based on context information that is acquired from mobile devices. Previous codebook learning methods are mainly developed on nonmobile platforms such as desktop PC, hence underutilize context features such as location and direction information as provided by the mobile devices. The proposed technique employs both the direction and location information to learn the codebook for mobile landmark recognition. A set of direction-aware leaf codewords are first generated by using direction data to decompose the leaf nodes of the original SVT. A visual word significance learning algorithm is then developed by considering location information to generate a compact codebook for image encoding. Experiments on the NTU50Landmark database show that the proposed method can achieve good recognition performance in mobile landmark recognition.

It has been estimated that autoimmune diseases are among the top ten leading causes of death among women in all age groups up to 65 years. However, the detection of it by indirect immunofluorescence (IIF) image analysis depends heavily on the experience of the physicians. An accurate and automatic Computer Aided Diagnosis (CAD) system will help greatly for the classification of the Human Epithelial type 2 (HEp-2) cell images with little human intervention. In this paper we present an automatic HEp-2 cell image classification technique that exploits different spatial scaled image representation and sparse coding of SIFT features. Additionally, spatial max pooling of sparse coding at different scales is used to boost the classification performance. The proposed method is tested on the ICPR 2012 contest dataset and experiments show that it clearly outperforms state-of-the-art techniques in cell and image level as well as two intensity level images.

Clinical features found in brain CT scan images are widely used in traumatic brain injury (TBI) as indicators for Glasgow Outcome Scale (GOS) prediction. However, due to the lack of automated methods to measure and quantify the CT scan image features, the computerized prediction of GOS in TBI has not been well studied. This paper introduces an automated GOS prediction system for traumatic brain CT images. Different from most existing systems that perform the prognosis based on pre-processed data, our system directly works on brain CT scan images based on the image features. Our system can also be extended to large dataset with easy adaptation. For each new image of a CT scan series, our proposed system first makes use of sparse representation model that predicts the GOS of each CT image slice using Gabor features. Logistic regression, which integrates the GOS of each CT scan slice with a pre-trained model, is then applied to estimate the GOS score for the new case which contains multiple CT slices. Evaluation of the system has shown promising results in prediction of GOS of traumatic brain injury cases.

Recognition of characters in natural images is a challenging task due to the complex background, variations of text size and perspective distortion, etc. Traditional optical character recognition (OCR) engine cannot perform well on those unconstrained text images. A novel technique is proposed in this paper that makes use of convolutional cooccurrence histogram of oriented gradient (ConvCoHOG), which is more robust and discriminative than both the histogram of oriented gradient (HOG) and the co-occurrence histogram of oriented gradients (CoHOG). In the proposed technique, a more informative feature is constructed by exhaustively extracting features from every possible image patches within character images. Experiments on two public datasets including the ICDAr 2003 Robust Reading character dataset and the Street View Text (SVT) dataset, show that our proposed character recognition technique obtains superior performance compared with state-of-the-art techniques.

The segmentation of scene text from the image background has shown great importance in scene text recognition. In this paper, we propose a multi-level MSER technology that identifies the best-quality text candidates from a set of stable regions that are extracted from different color channel images. In order to identify the best-quality text candidates, a segmentation score is defined which exploits four measures to evaluate the text probability of each stable region including: 1) Stroke width that measures the small stroke width variation of the text, 2) Boundary curvature that measures the smoothness of the stable region boundary, 3) Character confidence that measures the likelihood of a stable region being text based on a pre-trained support vector classifier, 4) Color constancy that measures the global color consistency of each selected text candidate. Finally, the MSERs with the best segmentation score from each channel are combined to form the final segmentation. The proposed method is evaluated on the ICDAR2003 and SVT datasets and experiments show that it outperforms both popular document image binarization methods and state of the art scene text segmentation methods.

This paper presents a visual saliency modeling technique that is efficient and tolerant to the image scale variation. Different from existing approaches that rely on a large number of filters or complicated learning processes, the proposed technique computes saliency from image histograms. Several two-dimensional image co-occurrence histograms are used, which encode not only "how many" (occurrence) but also "where and how" (co-occurrence) image pixels are composed into a visual image, hence capturing the "unusualness" of an object or image region that is often perceived by either global "uncommonness" (i.e., low occurrence frequency) or local "discontinuity" with respect to the surrounding (i.e., low co-occurrence frequency). The proposed technique has a number of advantageous characteristics. It is fast and very easy to implement. At the same time, it involves minimal parameter tuning, requires no training, and is robust to image scale variation. Experiments on the AIM dataset show that a superior shuffled AUC (sAUC) of 0.7221 is obtained, which is higher than the state-of-the-art sAUC of 0.7187.

Document Image Binarization is a technique to segment text out from the background region of a document image, which is a challenging task due to high intensity variations of the document foreground and background. Recently, a series of document image binarization contests (DIBCOs) had been held that have drawn great research interest in this area. Several document binarization techniques have been proposed and achieve great performance on the contest datasets. However, those proposed techniques may not perform well on all kinds of degraded document images because it is difficult to design a classification method that correctly models the non-uniform degraded document background and text foreground simultaneously. In this paper, we propose a self learning classification framework that combines binary outputs of different binarization methods. The proposed framework makes used of the sparse representation to re-classify the document pixels and produces a better binary results. The experimental results on the recent DIBCO contests show the great performance and robustness of our proposed framework on different kinds of degraded document images.

Scene text recognition is a fundamental step in End-to-End applications where traditional optical character recognition (OCR) systems often fail to produce satisfactory results. This paper proposes a technique that uses co-occurrence histogram of oriented gradients (Co-HOG) to recognize the text in scenes. Compared with histogram of oriented gradients (HOG), Co-HOG is a more powerful tool that captures spatial distribution of neighboring orientation pairs instead of just a single gradient orientation. At the same time, it is more efficient compared with HOG and therefore more suitable for real-time applications. The proposed scene text recognition technique is evaluated on ICDAR2003 character dataset and Street View Text (SVT) dataset. Experiments show that the Co-HOG based technique clearly outperforms state-of-the-art techniques that use HOG, Scale Invariant Feature Transform (SIFT), and Maximally Stable Extremal Regions (MSER).

Text detection in videos is challenging due to low resolution and complex background of videos. Besides, an arbitrary orientation of scene text lines in video makes the problem more complex and challenging. This paper presents a new method that extracts text lines of any orientations based on gradient vector flow (GVF) and neighbor component grouping. The GVF of edge pixels in the Sobel edge map of the input frame is explored to identify the dominant edge pixels which represent text components. The method extracts edge components corresponding to dominant pixels in the Sobel edge map, which we call text candidates (TC) of the text lines. We propose two grouping schemes. The first finds nearest neighbors based on geometrical properties of TC to group broken segments and neighboring characters which results in word patches. The end and junction points of skeleton of the word patches are considered to eliminate false positives, which output the candidate text components (CTC). The second is based on the direction and the size of the CTC to extract neighboring CTC and to restore missing CTC, which enables arbitrarily oriented text line detection in video frame. Experimental results on different datasets, including arbitrarily oriented text data, nonhorizontal and horizontal text data, Hua's data and ICDAR-03 data (camera images), show that the proposed method outperforms existing methods in terms of recall, precision and f-measure.

Segmentation of text from badly degraded document images is a very challenging task due to the high inter/intra-variation between the document background and the foreground text of different document images. In this paper, we propose a novel document image binarization technique that addresses these issues by using adaptive image contrast. The adaptive image contrast is a combination of the local image contrast and the local image gradient that is tolerant to text and background variation caused by different types of document degradations. In the proposed technique, an adaptive contrast map is first constructed for an input degraded document image. The contrast map is then binarized and combined with Canny's edge map to identify the text stroke edge pixels. The document text is further segmented by a local threshold that is estimated based on the intensities of detected text stroke edge pixels within a local window. The proposed method is simple, robust, and involves minimum parameter tuning. It has been tested on three public datasets that are used in the recent document image binarization contest (DIBCO) 2009 & 2011 and handwritten-DIBCO 2010 and achieves accuracies of 93.5%, 87.8%, and 92.03%, respectively, that are significantly higher than or close to that of the best-performing methods reported in the three contests. Experiments on the Bickley diary dataset that consists of several challenging bad quality document images also show the superior performance of our proposed method, compared with other techniques.

Musical staff line detection and removal techniques detect the staff positions in musical documents and segment musical score from musical documents by removing those staff lines. It is an important preprocessing step for ensuing the Optical Music Recognition tasks. This paper proposes an effective staff line detection and removal method that makes use of the global information of the musical document and models the staff line shape. It first estimates the staff height and space, and then models the shape of the staff line by examining the orientation of the staff pixels. At last the estimated model is used to find out the location of staff lines and hence to remove those detected staff lines. The proposed technique is simple, robust, and involves few parameters. It has been tested on the dataset of the recent staff removal competition held under the International Conference of Document Analysis and Recognition(ICDAR) 2011. Experimental results show the effectiveness and robustness of our proposed technique on musical documents with various types of deformations.

In this paper, we present new features based on Spatial-Gradient-Features (SGF) at block level for identifying six video scripts namely, Arabic, Chinese, English, Japanese, Korean and Tamil. This works helps in enhancing the capability of the current OCR on video text recognition by choosing an appropriate OCR engine when video contains multi-script frames. The input for script identification is the text blocks obtained by our text frame classification method. For each text block, we obtain horizontal and vertical gradient information to enhance the contrast of the text pixels. We divide the horizontal gradient block into two equal parts as upper and lower at the centroid in the horizontal direction. Histogram on the horizontal gradient values of the upper and the lower part is performed to select dominant text pixels. In the same way, the method selects dominant pixels from the right and the left parts obtained by dividing the vertical gradient block vertically. The method combines the horizontal and the vertical dominant pixels to obtain text components. Skeleton concept is used to reduce pixel width to a single pixel to extract spatial features. We extract four features based on proximity between end points, junction points, intersection points and pixels. The method is evaluated on 770 frames of six scripts in terms of classification rate and is compared with an existing method. We have achieved 82.1% average classification rate.

Multioriented text detection in video frames is not as easy as detection of captions or graphics or overlaid texts, which usually appears in the horizontal direction and has high contrast compared to its background. Multioriented text generally refers to scene text that makes text detection more challenging and interesting due to unfavorable characteristics of scene text. Therefore, conventional text detection methods may not give good results for multioriented scene text detection. Hence, in this paper, we present a new enhancement method that includes the product of Laplacian and Sobel operations to enhance text pixels in videos. To classify true text pixels, we propose a Bayesian classifier without assuming a priori probability about the input frame but estimating it based on three probable matrices. Three different ways of clustering are performed on the output of the enhancement method to obtain the three probable matrices. Text candidates are obtained by intersecting the output of the Bayesian classifier with the Canny edge map of the input frame. A boundary growing method is introduced to traverse the multioriented scene text lines using text candidates. The boundary growing method works based on the concept of nearest neighbors. The robustness of the method has been tested on a variety of datasets that include our own created data (nonhorizontal and horizontal text data) and two publicly available data, namely, video frames of Hua and complex scene text data of ICDAR 2003 competition (camera images). Experimental results show that the performance of the proposed method is encouraging compared with results of existing methods in terms of recall, precision, F-measures, and computational times.

In this paper, we present a new method for video script identification which is essential before choosing an appropriate OCR engine for identifying text lines when a video frame contains more than one language. The input for script identification is the text lines obtained by our text detection method. We extract upper and lower extreme points for each connected component of Canny edges of text lines. The extracted points are connected to study the behavior of upper and lower lines. The direction of each 10-pixel segment of the lines is determined using PCA. The average angle of the segments of the upper and lower lines is computed to study the smoothness and cursiveness of the lines. In addition, to discriminate the scripts accurately, the method divides a text line into five equal zones horizontally to study the smoothness and cursiveness of the upper and lower lines of each zone. We evaluate the method by conducting experiments on different combinations of languages such as English and Chinese, English and Tamil, Chinese and Tamil, and English, Chinese and Tamil.

This paper presents a new method based on Fourier and moments features to extract words and characters from a video text line in any direction for recognition. Unlike existing methods which output the entire text line to the ensuing recognition algorithm, the proposed method obtains each extracted character from the text line as input to the recognition algorithm because the background of a single character is relatively simple compared to the text line and words. Max-Min clustering criterion is introduced to obtain text cluster from the extracted Fourier and moments feature set. Union of the text cluster with Canny operation of the input video text line is proposed to obtain missing text candidates. Then a run length criterion is used for extraction of words. From the words, we propose a new idea for extracting characters from the text candidates of each word image based on the fact that the text height difference at the character boundary column is smaller than that at other columns of the word image. We evaluate the method on a large dataset at three levels namely text line, words and characters in terms of recall, precision and f-measure. In addition to this, we show that the recognition result for the extracted character is better than words and lines. Our experimental set up involves 3527 characters including Chinese. The dataset is selected from TRECVID database of 2005 and 2006.

We present a new video character recognition method based on hierarchical classification. In the first step, we propose a method for character segmentation of the text line detected by the text detection method. The segmentation algorithm uses dynamic programming to find least-cost paths in the gray domain to identify the spaces between characters. For the segmented characters, we get a Canny edge image as input for the character recognition step. We introduce hierarchical classification based on voting criteria with structural features to classify 62 character classes into different smaller classes. We divide the perimeter of a character into 8 segments according to 8 directions at the centroid. Then the shape of each segment is studied to recognize the characters based on distances between the centroid and end points, and distances between the midpoint and end points. Our experiments on 1462 characters of upper case, lower case and numerals shows that 10% samples per class for training is enough to obtain 94.5% recognition accuracy. The dataset is chosen from TRECVID database of 2005 and 2006.

Document image binarization has been studied for decades, and many practical binarization techniques have been proposed for different kinds of document images. However, many state-of-the-art methods are particularly suitable for the document images that suffer from certain specific type of image degradation or have certain specific type of image characteristics. In this paper, we propose a classification framework to combine different thresholding methods and produce better performance for document image binarization. Given the binarization results of some reported methods, the proposed framework divides the document image pixels into three sets, namely, foreground pixels, background pixels and uncertain pixels. A classifier is then applied to iteratively classify those uncertain pixels into foreground and background, based on the pre-selected froeground and background sets. Extensive experiments over different datasets including the Document Image Binarization Contest(DIBCO)2009 and Handwritten Document Image Binarization Competition(H-DIBCO)2010 show that our proposed framework outperforms most state-of-the-art methods significantly.

Many digital images contain blurred regions which are caused by motion or defocus. Automatic detection and classification of blurred image regions are very important for different multimedia analyzing tasks. This paper presents a simple and effective automatic image blurred region detection and classification technique. In the proposed technique, blurred image regions are first detected by examining singular value information for each image pixels. The blur types (i.e. motion blur or defocus blur) are then determined based on certain alpha channel constraint that requires neither image deblurring nor blur kernel estimation. Extensive experiments have been conducted over a dataset that consists of 200 blurred image regions and 200 image regions with no blur that are extracted from 100 digital images. Experimental results show that the proposed technique detects and classifies the two types of image blurs accurately. The proposed technique can be used in many different multimedia analysis applications such as image segmentation, depth estimation and information retrieval.

Under the framework of computer-aided diagnosis, this paper presents an accurate and efficient optic disc (OD) detection and segmentation technique. A circular transformation is designed to capture both the circular shape of the OD and the image variation across the OD boundary simultaneously. For each retinal image pixel, it evaluates the image variation along multiple evenly-oriented radial line segments of specific length. The pixels with the maximum variation along all radial line segments are determined, which can be further exploited to locate both the OD center and the OD boundary accurately. Experiments show that OD detection accuracies of 99.75%, 97.5%, and 98.77% are obtained for the STARE dataset, the ARIA dataset, and the MESSIDOR dataset, respectively, and the OD center error lies around six pixels for the STARE dataset and the ARIA dataset which is much smaller than that of state-of-the-art methods ranging 14-29 pixels. In addition, the OD segmentation accuracies of 93.4% and 91.7% are obtained for STARE dataset and ARIA dataset, respectively, that consists of many severely degraded images of pathological retinas that state-of-the-art methods cannot segment properly. Furthermore, the algorithm runs in 5 s, which is substantially faster than many of the state-of-the-art methods.

This paper presents a new document image binarization technique that segments the text from badly degraded historical document images. The proposed technique makes use of the image contrast that is defined by the local image maximum and minimum. Compared with the image gradient, the image contrast evaluated by the local maximum and minimum has a nice property that it is more tolerant to the uneven illumination and other types of document degradation such as smear. Given a historical document image, the proposed technique first constructs a contrast image and then detects the high contrast image pixels which usually lie around the text stroke boundary. The document text is then segmented by using local thresholds that are estimated from the detected high contrast pixels within a local neighborhood window. The proposed technique has been tested over the dataset that is used in the recent Document Image Binarization Contest (DIBCO) 2009. Experiments show its superior performance.

This paper presents an automatic optic disc (OD) detection technique. Given a retinal image, the proposed method first estimates a retinal background surface through an iterative Savitzky-Golay smoothing procedure. The OD is then detected through the global thresholding of the difference between the retinal image and the estimated background surface. Finally, an OD boundary is determined after a pair of morphological post-processing operations. The proposed technique has been tested over three public datasets that are composed of 130, 89, and 40 retinal images, respectively. Experiments show that an average OD detection accuracy of 96.91% is attained. In addition, 84.37% OD pixels are correctly located compared with the manually labeled ones.

This paper presents an automatic macula detection technique that makes use of the circular brightness profile of the macula: the macula is usually darker than the surrounding pixels whose intensities increase gradually with their distances from the macula center. A line operator is designed to capture the macula circular brightness profile, which evaluates the image brightness variation along multiple line segments of specific orientations that pass through each retinal image pixel. The orientation of the line segment with the minimum/ maximum variation has specific patterns that indicate the position of the macula efficiently. The proposed technique has been tested over DRIVE project's dataset and the STARE project's dataset. Experiments show that the accuracies reach up to 100% and 95.45%, respectively, based on 35 and 44 retinal images having discernible macula within the two public datasets.

Document Image Binarization techniques have been studied for many years, and many practical binarization techniques have been developed and applied successfully on commercial document analysis systems. However, the current state-of-the-art methods, fail to produce good binarization results for many badly degraded document images. In this paper, we propose a self-training learning framework for document image binarization. Based on reported binarization methods, the proposed framework first divides document image pixels into three categories, namely, foreground pixels, background pixels and uncertain pixels. A classifier is then trained by learning from the document image pixels in the foreground and background categories. Finally, the uncertain pixels are classified using the learned pixel classifier. Extensive experiments have been conducted over the dataset that is used in the recent Document Image Binarization Contest (DIBCO) 2009. Experimental results show that our proposed framework significantly improves the performance of reported document image binarization methods.

In the process of designing tailored information delivery systems based on natural language generation technology, linguistic resources need to be developed. One of such resources is a set of discourse strategies, the abstract structures that specify how to construct a coherent text to achieve a specific communicative goal. Discourse strategies, however, prove to be difficult to develop, in part due to their abstractness. In this paper, we describe a graphical tool which allows users to annotate documents to indicate its various elements and how they relate to each other. The annotations can then be used to generate a first draft of the discourse strategies automatically. The tool is presented in the context of the Myriad delivery platform which aims at facilitating the rapid development of tailored information delivery systems.

This paper presents a photometric restoration technique that automatically corrects shading within retinal images taken with a fundus camera. The proposed technique is based on the observation that the background of retinal images usually shows flat reflectance variations due to its high similarity in color and texture. It estimates shading through an iterative polynomial interpolation procedure that first estimates a shading image through a horizontal interpolation process and then improves the shading estimation by a vertical interpolation process. Once the shading image is estimated, a reflectance image can accordingly be determined based on the luminance of the retina image under study. Experiments on 161 retinal images of different qualities show promising results.

This paper presents a subject-independent EEG (Electroencephalogram) classification technique and its application to a P300-based word speller. Due to EEG variations across subjects, a user calibration procedure is usually required to build a subject-specific classification model (SSCM). We remove the user calibration through the boosting of a committee of weak classifiers learned from EEG of a pool of subjects. In particular, we ensemble the weak classifiers based on their confidence that is evaluated according to the classification consistency. Experiments over ten subjects show that the proposed technique greatly outperforms the supervised classification models, hence making P300-based BCIs more convenient for practical uses.

This paper proposes an approach to learn subject-independent P300 models for EEG-based brain-computer interfaces. The P300 models are first learned using a pool of existing subjects and Fisher linear discriminant, and then autonomously adapted to the unlabeled data of a new subject using an unsupervised machine learning technique. In data analysis, we apply this technique to a set of EEG data of 10 subjects performing word spelling in an oddball paradigm. The results are very positive: the adapted models with unlabeled data yield virtually the same classification accuracy as the conventional methods with labeled data. Therefore, it proves the feasibility of P300-based BCIs which can be applied directly to a new subject without training sessions.

This paper presents a document retrieval technique that is capable of searching document images without OCR (optical character recognition). The proposed technique retrieves document images by a new word shape coding scheme, which captures the document content through annotating each word image by a word shape code. In particular, we annotate word images by using a set of topological shape features including character ascenders/descenders, character holes, and character water reservoirs. With the annotated word shape codes, document images can be retrieved by either query keywords or a query document image. Experimental results show that the proposed document image retrieval technique is fast, efficient, and tolerant to various types of document degradation.

This paper reports an identification technique that detects scripts and languages of noisy and degraded document images. In the proposed technique, scripts and languages are identified through the document vectorization, which converts each document image into a document vector that characterizes the shape and frequency of the conta ned character or word images. Document images are vectorized by using vertical component cuts and character extremum points, which are both tolerant to the variation in text fonts and styles, noise, and various types of document degradation. For each script or language under study, a script or language template is first constructed through a training process. Scripts and languages of document images are then determined according to the distances between converted document vectors and the pre-constructed script and language templates. Experimental results show that the proposed technique is accurate, easy for extension, and tolerant to noise and various types of document degradation.

This paper presents a document image thresholding technique that binarizes badly illuminated document images by the photometric correction. Based on the observation that illumination normally varies smoothly and document images often contain a uniformly colored background, the global shading variation is estimated by using a two-dimensional Savitzky-Golay filter that fits a least square polynomial surface to the luminance of a badly illuminated document image. With the knowledge of the global shading variation, shading degradation is then corrected through a compensation process that produces animage with roughly uniform illumination. Badly illuminated document images are accordingly binarized through the global thresholding of the compensated ones. Experiments show that the proposed thresholding technique is fast, robust, and efficient for the binarization of badly illuminated document images.

In order to capture the content of an imaged document but avoid the time-consuming full-scale OCR which is fragile to handle touching characters, a fast and segmentation- free keyword spotting method is proposed in this paper. The keyword spotting method is based on word shape coding technique. The proposed coding scheme has little ambiguity, and can be swiftly executed. It is a promising technique to boost better document image retrieval. The strength of the proposed method is demonstrated in a document filtering experiment. The experimental results show that document filtering based on the proposed method is more than 20 times faster than the one based on OCR, and has comparable filtering accuracy.

This paper presents a language identification technique that detects Latin-based languages of imaged documents without OCR. The proposed technique detects languages through the word shape coding, which converts each word image into a word shape code and accordingly transforms each document image into an electronic document vector. For each Latin-based language under study, a language template is first constructed through a corpus-based learning process. The underlying language of the query document is then determined based on the similarity between the query document vector and multiple constructed language templates. Compared with the reported methods, the proposed language identification technique is fast, accurate, and tolerant to text segmentation error caused by noise and various types of document degradation. Experimental results show some promising results.

This paper presents an identification technique that automatically detects the underlying script and orientation of scanned document images. In the proposed technique, document script and orientation are identified by using the stroke density and distribution, which convert each document image into a document vector. For each script at each orientation, a number of reference document vectors are first constructed. Script and orientation of the query document are then determined according to the similarity between the query document vector and multiple pre- constructed reference document vectors by using the K-nearest neighbor algorithm. Experiments show that the proposed technique is tolerant to the document skew and able to detect orientations of documents of different scripts.

This paper presents a document image binarization technique that segments text from badly illuminated document images. Based on the observations that text documents normally lie over a planar or smoothly curved surface and have a uniformly colored background, badly illuminated document images are binarized by using a smoothing polynomial surface, which estimates the shading variation and compensates the shading degradation based on the estimated shading variation. Badly illuminated document images are accordingly binarized through the global thresholding of the compensated document images. Compared with the reported methods, the proposed technique is tolerant to the variations in text size and document contrast. At the same time, it is much faster and able to produce a binary text image with little background noise.

This paper presents a keyword spotting technique that locates keywords within document images captured by a digital camera. In the proposed technique, the shape of word images in perspective view is captured by using three perspective invariants, namely, holes, water reservoirs, and character ascenders and descenders. Given a camera image of document, text line and word images are first segmented through the connected component analysis. The three perspective invariants are then detected through two rounds of scanning process, which transliterate each character image into a character shape code of dimension six and so convert each word image into a word shape code. Keywords within camera images of documents are finally located through a partial matching process. Experiments show some promising results.

This paper presents a document skew and orientation detection technique. The proposed technique estimates document skew and orientation based on the observation that text images normally hold a large amount of equidistant interline spacings and the number of character ascenders is statistically much larger than that of character descenders. Given a document image with arbitrary skew and orientation, white run histograms are first constructed through scanning documents in horizontal and vertical directions. Document skew is then estimated by using the white runs that exactly span the interline spacing. Lastly, document orientation is determined according to the numbers of character ascenders and descenders, which are detected by using the white runs that cross the interline spacing and lie over character ascenders and descenders. Experiments show that the proposed technique is fast, accurate, and capable of detecting arbitrary document skew and orientation.

This paper presents an automatic orientation detection and categorization technique that is capable of detecting the orientation of multilingual documents with arbitrary skew and categorizing document images according to the underlying languages. We carry out orientation detection and categorization through document vectorization, which encodes document orientation and language information and converts each document image into an electronic document vector through the exploitation of the density and distribution of vertical component runs. For each language of interest, a pair of vector templates is first constructed through a training process. Orientation and category of the query image are then determined based on distances between the query document vector and the constructed vector templates. Experiments over 492 testing document images show that the average orientation detection and categorization rates reach up to 97.56% and 99.59%, respectively.

In this paper, we deal with the problem of document image rectification from images captured by digital cameras. The improvement on the resolution of digital camera sensors has brought more and more applications for non-contact text capture. Unfortunately, perspective distortion coupled with resulting images makes it harder to properly identify the contents of captured texts using the traditional optical character recognition (OCR) system. We propose in this work a new technique, which is capable of removing distortion and recovering the fronto-parallel view of text with a single image. Different from reported approaches in the literature, the image rectification is carried out using character boundary and tip point, which are extracted from character strokes based on multiple fuzzy sets and morphological operators. The algorithm needs neither camera calibration nor high-contrast document boundary. Experimental results show our rectification process is fast and robust.