Existing retrieval-based methods have made significant strides in maintaining long-term conversations. However, these approaches face challenges in memory database management and accurate memory retrieval, hindering their efficacy in dynamic, real-world interactions. This study introduces a novel framework, COmpressive Memory-Enhanced Dialogue sYstems (COMEDY), which eschews traditional retrieval modules and memory databases. Instead, COMEDY adopts a “One-for-All” approach, utilizing a single language model to manage memory generation, compression, and response generation. Central to this framework is the concept of compressive memory, which integrates session-specific summaries, user-bot dynamics, and past events into a concise memory format. To support COMEDY, we collect the biggest Chinese long-term conversation dataset, Dolphin, derived from real user-chatbot interactions. Comparative evaluations demonstrate COMEDY’s superiority over traditional retrieval-based methods in producing more nuanced and human-like conversational experiences.

Generating natural listener responses in conversational scenarios is crucial for creating engaging digital humans and avatars. Recent work has shown that large language models (LLMs) can be effectively leveraged for this task, demonstrating remarkable capabilities in generating contextually appropriate listener behaviors. However, current LLM-based methods face two critical limitations: they rely solely on speech content, overlooking other crucial communication signals, and they entangle listener identity with response generation, compromising output fidelity and generalization. In this work, we present a novel framework that addresses these limitations while maintaining the advantages of LLMs. Our approach introduces a Multimodal-LM architecture that jointly processes speech content, acoustics, and speaker emotion, capturing the full spectrum of communication cues. Additionally, we propose an identity disentanglement strategy using instance normalization and adaptive instance normalization in a VQ-VAE framework, enabling high-fidelity listening head synthesis with flexible identity control. Extensive experiments demonstrate that our method significantly outperforms existing approaches in terms of response naturalness and fidelity, while enabling effective identity control without retraining.

Patch-based image tokenization ignores the morphology of the visual world, limiting effective and efficient learning of image understanding. Inspired by subword tokenization, we introduce subobject-level adaptive token segmentation and explore several approaches, including superpixel, SAM, and a proposed Efficient and PanOptiC (EPOC) image tokenizer. Our EPOC combines boundary detection–a simple task that can be handled well by a compact model–with watershed segmentation, which inherently guarantees no pixels are left unsegmented. Intrinsic evaluations across 5 datasets demonstrate that EPOC’s segmentation aligns well with human annotations of both object- and part-level visual morphology, producing more monosemantic tokens and offering substantial efficiency advantages. For extrinsic evaluation, we designed a token embedding that handles arbitrary-shaped tokens, and trained VLMs with different tokenizers on 4 datasets of object recognition and detailed captioning. The results reveal that subobject tokenization enables faster convergence and better generalization while using fewer visual tokens.

Although existing speech-driven talking face generation methods achieve significant progress, they are far from real-world application due to the avatar-specific training demand and unstable lip movements. To address the above issues, we propose the GSmoothFace, a novel two-stage generalized talking face generation model guided by a fine-grained 3D face model, which can synthesize smooth lip dynamics while preserving the speaker's identity. Our proposed GSmoothFace model mainly consists of the Audio to Expression Prediction (A2EP) module and the Target Adaptive Face Translation (TAFT) module. Specifically, we first develop the A2EP module to predict expression parameters synchronized with the driven speech. It uses a transformer to capture the long-term audio context and learns the parameters from the fine-grained 3D facial vertices, resulting in accurate and smooth lip-synchronization performance. Afterward, the well-designed TAFT module, empowered by Morphology Augmented Face Blending (MAFB), takes the predicted expression parameters and target video as inputs to modify the facial region of the target video without distorting the background content. The TAFT effectively exploits the identity appearance and background context in the target video, which makes it possible to generalize to different speakers without retraining. Both quantitative and qualitative experiments confirm the superiority of our method in terms of realism, lip-synchronization, and visual quality.

Recent research has demonstrated that the multi-task fine-tuning of multi-modal Large Language Models (LLMs) using an assortment of annotated downstream vision-language datasets significantly enhances their performance. Yet, during this process, a side effect, which we termed as the "multi-modal alignment tax", surfaces. This side effect negatively impacts the model's ability to format responses appropriately - for instance, its "politeness" - due to the overly succinct and unformatted nature of raw annotations, resulting in reduced human preference. In this paper, we introduce Polite Flamingo, a multi-modal response rewriter that transforms raw annotations into a more appealing, "polite" format. Polite Flamingo is trained to reconstruct high-quality responses from their automatically distorted counterparts and is subsequently applied to a vast array of vision-language datasets for response rewriting. After rigorous filtering, we generate the PF-1M dataset and further validate its value by fine-tuning a multi-modal LLM with it. Combined with novel methodologies including U-shaped multi-stage tuning and multi-turn augmentation, the resulting model, Clever Flamingo, demonstrates its advantages in both multi-modal understanding and response politeness according to automated and human evaluations. Code and dataset are available at https://github.com/ChenDelong1999/polite-flamingo

Existing one-shot 4D head synthesis methods usually learn from monocular videos with the aid of 3DMM re-construction, yet the latter is evenly challenging which re-stricts them from reasonable 4D head synthesis. We present a method to learn one-shot 4D head synthesis via large-scale synthetic data. The key is to first learn a part-wise 4D generative model from monocular images via adver-sarial learning, to synthesize multi-view images of diverse identities and full motions as training data; then leverage a transformer-based animatable triplane reconstructor to learn 4D head reconstruction using the synthetic data. A novel learning strategy is enforced to enhance the general-izability to real images by disentangling the learning pro-cess of 3D reconstruction and reenactment. Experiments demonstrate our superiority over the prior art.

In this paper, we propose a novel virtual try-on from unconstrained designs (ucVTON) task to enable photorealistic synthesis of personalized composite clothing on input human images. Unlike prior arts constrained by specific input types, our method allows flexible specification of style (text or image) and texture (full garment, cropped sections, or texture patches) conditions. To address the entanglement challenge when using full garment images as conditions, we develop a two-stage pipeline with explicit disentanglement of style and texture. In the first stage, we generate a human parsing map reflecting the desired style conditioned on the input. In the second stage, we composite textures onto the parsing map areas based on the texture input. To represent complex and non-stationary textures that have never been achieved in previous fashion editing works, we first propose extracting hierarchical and balanced CLIP features and applying position encoding in VTON. Experiments demonstrate superior synthesis quality and personalization enabled by our method. The flexible control over style and texture mixing brings virtual try-on to a new level of user experience for online shopping and fashion design.

As for human avatar reconstruction, contemporary techniques commonly necessitate the acquisition of costly data and struggle to achieve satisfactory results from a small number of casual images. In this paper, we investigate this task from a few-shot unconstrained photo album. The reconstruction of human avatars from such data sources is challenging because of limited data amount and dynamic articulated poses. For handling dynamic data, we integrate a skinning mechanism with deep marching tetrahedra (DMTet) to form a drivable tetrahedral representation, which drives arbitrary mesh topologies generated by the DMTet for the adaptation of unconstrained images. To effectively mine instructive information from few-shot data, we devise a two-phase optimization method with few-shot reference and few-shot guidance. The former focuses on aligning avatar identity with reference images, while the latter aims to generate plausible appearances for unseen regions. Overall, our framework, called HaveFun, can undertake avatar reconstruction, rendering, and animation. Extensive experiments on our developed benchmarks demonstrate that HaveFun exhibits substantially superior performance in reconstructing the human body and hand.

Large Language Models(LLMs) have shown remarkable emergent abilities in unifying almost all (if not every) NLP tasks. In the human motion-related realm, however, researchers still develop siloed models for each task. In-spired by InstuctGPT[16], and the generalist concept be-hind Gato [27], we introduce AvatarGPT, an All-in-One framework for motion understanding, planning, generations as well as other tasks such as motion in-between synthesis. AvatarGPT treats each task as one type of in-struction fine-tuned on the shared LLM. All the tasks are seamlessly interconnected with language as the univer-sal interface, constituting a closed-loop within the frame-work. To achieve this, human motion sequences are first encoded as discrete tokens, which serve as the extended vo-cabulary of LLM. Then, an unsupervised pipeline to gen-erate natural language descriptions of human action sequences from in-the-wild videos is developed. Finally, all tasks are jointly trained. Extensive experiments show that AvatarGPT achieves SOTA on low-level tasks, and promising results on high-level tasks, demonstrating the effectiveness of our proposed All-in-One framework. Moreover, for the first time, AvatarGPT enables a principled approach by iterative traversal of the tasks within the closed-loop for un-limited long-motion synthesis.

How to generate diverse, life-like, and unlimited long head/body sequences without any driving source? We argue that this under-investigated research problem is nontrivial at all, and has unique technical challenges behind it. Without semantic constraints from the driving sources, using the standard autoregressive model to generate infinitely long sequences would easily result in 1) out-of-distribution (OOD) issue due to the accumulated error, 2) insufficient diversity to produce natural and life-like motion sequences and 3) undesired periodic patterns along the time. To tackle the above challenges, we propose a systematic framework that marries the benefits of VQ-VAE and a novel token-level control policy trained with reinforcement learning using carefully designed reward functions. A high-level prior model can be easily injected on top to generate unlimited long and diverse sequences. Although we focus on no driving sources now, our framework can be generalized for controlled synthesis with explicit driving sources. Through comprehensive evaluations, we conclude that our proposed framework can address all the above-mentioned challenges and outperform other strong baselines very significantly.

3D single object tracking in LiDAR point clouds (LiDAR SOT) plays a crucial role in autonomous driving. Current approaches all follow the Siamese paradigm based on appearance matching. However, LiDAR point clouds are usually textureless and incomplete, which hinders effective appearance matching. Besides, previous methods greatly overlook the critical motion clues among targets. In this work, beyond 3D Siamese tracking, we introduce amotion-centric paradigmto handle LiDAR SOT from a new perspective. Following this paradigm, we propose a matching-free two-stage trackerM$^{2}$2-Track. At the 1st-stage,$M^{2}$-Track localizes the target within successive frames viamotion transformation. Then it refines the target box throughmotion-assisted shape completion at the 2nd-stage. Due to the motion-centric nature, our method shows its impressive generalizability with limited training labels and provides good differentiability for end-to-end cycle training. This inspires us to explore semi-supervised LiDAR SOT by incorporating a pseudo-label-based motion augmentation and a self-supervised loss term. Under the fully-supervised setting, extensive experiments confirm that$M^{2}$-Track significantly outperforms previous state-of-the-arts on three large-scale datasets while running at57FPS($\sim$∼3%,$\sim$∼11%and$\sim$∼22%precision gains on KITTI, NuScenes, and Waymo Open Dataset respectively). While under the semi-supervised setting, our method performs on par with or even surpasses its fully-supervised counterpart using fewer than half labels from KITTI. Further analysis verifies each component's effectiveness and shows the motion-centric paradigm's promising potential for auto-labeling and unsupervised domain adaptation.

Open-domain dialogue systems have made promising progress in recent years.While the state-of-the-art dialogue agents are built upon large-scale text-based social media data and large pre-trained models, there is no guarantee these agents could also perform well in fast-growing scenarios, such as live streaming, due to the bounded transferability of pretrained models and biased distributions of public datasets from Reddit and Weibo, etc.To improve the essential capability of responding and establish a benchmark in the live opendomain scenario, we introduce the LiveChat dataset, composed of 1.33 million real-life Chinese dialogues with almost 3800 average sessions across 351 personas and fine-grained profiles for each persona.LiveChat is automatically constructed by processing numerous live videos on the Internet and naturally falls within the scope of multi-party conversations, where the issues of Who says What to Whom should be considered.Therefore, we target two critical tasks of response modeling and addressee recognition and propose retrieval-based baselines grounded on advanced techniques.Experimental results have validated the positive effects of leveraging persona profiles and larger average sessions per persona.In addition, we also benchmark the transferability of advanced generation-based models on LiveChat and pose some future directions for current challenges. 1

Due to the complex label hierarchy and intensive labeling cost in practice, the hierarchical text classification (HTC) suffers a poor performance especially when low-resource or fewshot settings are considered.Recently, there is a growing trend of applying prompts on pretrained language models (PLMs), which has exhibited effectiveness in the few-shot flat text classification tasks.However, limited work has studied the paradigm of prompt-based learning in the HTC problem when the training data is extremely scarce.In this work, we define a path-based few-shot setting and establish a strict path-based evaluation metric to further explore few-shot HTC tasks.To address the issue, we propose the hierarchical verbalizer ("Hi-erVerb"), a multi-verbalizer framework treating HTC as a single-or multi-label classification problem at multiple layers and learning vectors as verbalizers constrained by hierarchical structure and hierarchical contrastive learning.In this manner, HierVerb fuses label hierarchy knowledge into verbalizers and remarkably outperforms those who inject hierarchy through graph encoders, maximizing the benefits of PLMs.Extensive experiments on three popular HTC datasets under the few-shot settings demonstrate that prompt with HierVerb significantly boosts the HTC performance, meanwhile indicating an elegant way to bridge the gap between the large pre-trained model and downstream hierarchical classification tasks. 1

We present HandAvatar, a novel representation for hand animation and rendering, which can generate smoothly compositional geometry and self-occlusion-aware texture. Specifically, we first develop a MANO-HD model as a high-resolution mesh topology to fit personalized hand shapes. Sequentially, we decompose hand geometry into per-bone rigid parts, and then re-compose paired geometry encodings to derive an across-part consistent occupancy field. As for texture modeling, we propose a self-occlusion-aware shading field (SelF). In SelF, drivable anchors are paved on the MANO-HD surface to record albedo information under a wide variety of hand poses. Moreover, directed soft occupancy is designed to describe the ray-to-surface relation, which is leveraged to generate an illumination field for the disentanglement of pose-independent albedo and pose-dependent illumination. Trained from monocular video data, our HandAvatar can perform freepose hand animation and rendering while at the same time achieving superior appearance fidelity. We also demonstrate that HandAvatar provides a route for hand appearance editing. Project website: https://seanchenxy.github.iO/HandAvatarWeb.

A key challenge for novel view synthesis of monocular portrait images is 3D consistency under continuous pose variations. Most existing methods rely on 2D generative models which often leads to obvious 3D inconsistency artifacts. We present a 3D-consistent novel view synthesis approach for monocular portrait images based on a recent proposed 3D-aware GAN, namely Generative Radiance Manifolds (GRAM) [13], which has shown strong 3D consistency at multiview image generation of virtual subjects via the radiance manifolds representation. However, simply learning an encoder to map a real image into the latent space of GRAM can only reconstruct coarse radiance manifolds without faithful fine details, while improving the reconstruction fidelity via instance-specific optimization is time-consuming. We introduce a novel detail manifolds reconstructor to learn 3D-consistent fine details on the radiance manifolds from monocular images, and combine them with the coarse radiance manifolds for high-fidelity reconstruction. The 3D priors derived from the coarse radiance manifolds are used to regulate the learned details to ensure reasonable synthesized results at novel views. Trained on in-the-wild 2D images, our method achieves high-fidelity and 3D-consistent portrait synthesis largely outperforming the prior art. Project page: https://yudeng.github.io/GRAMInverter/

We present a novel one-shot talking head synthesis method that achieves disentangled and fine-grained control over lip motion, eye gaze&blink, head pose, and emotional expression. We represent different motions via disentangled latent representations and leverage an image generator to synthesize talking heads from them. To effectively disentangle each motion factor, we propose a progressive disentangled representation learning strategy by separating the factors in a coarse-to-fine manner, where we first extract unified motion feature from the driving signal, and then isolate each fine-grained motion from the unified feature. We leverage motion-specific contrastive learning and regressing for non-emotional motions, and introduce feature-level decorrelation and self-reconstruction for emotional expression, to fully utilize the inherent properties of each motion factor in unstructured video data to achieve disentanglement. Experiments show that our method provides high quality speech&lip-motion synchronization along with precise and disentangled control over multiple extra facial motions, which can hardly be achieved by previous methods.

Generating controllable and editable human motion sequences is a key challenge in 3D Avatar generation. It has been labor-intensive to generate and animate human motion for a long time until learning-based approaches have been developed and applied recently. However, these approaches are still task-specific or modality-specific [1][6] [5][18]. In this paper, we propose “UDE”, the first unified driving engine that enables generating human motion sequences from natural language or audio sequences (see Fig. 1). Specifically, UDE consists of the following key components: 1) a motion quantization module based on VQVAE that represents continuous motion sequence as discrete latent code [33], 2) a modality-agnostic transformer encoder [34] that learns to map modality-aware driving signals to a joint space, and 3) a unified token transformer (GPT-like[24])network to predict the quantized latent code index in an auto-regressive manner. 4) a diffusion motion decoder that takes as input the motion tokens and decodes them into motion sequence swith high diversity. Weevaluate our method on HumanML3D [8] and AIST+ [19] benchmarks, and the experiment results demonstrate our method achieves state-of-the-art performance. Project website: https://zixiangzhou916.github.io/UDE/

Chain-of-Thought (CoT) prompting has proven to be effective in enhancing the reasoning capabilities of Large Language Models (LLMs) with at least 100 billion parameters.However, it is ineffective or even detrimental when applied to reasoning tasks in Smaller Language Models (SLMs) with less than 10 billion parameters.To address this limitation, we introduce Dialogue-guided Chain-of-Thought (DialCoT) which employs a dialogue format to generate intermediate reasoning steps, guiding the model toward the final answer.Additionally, we optimize the model's reasoning path selection using the Proximal Policy Optimization (PPO) algorithm, further enhancing its reasoning capabilities.Our method offers several advantages compared to previous approaches.Firstly, we transform the process of solving complex reasoning questions by breaking them down into a series of simpler sub-questions, significantly reducing the task difficulty and making it more suitable for SLMs.Secondly, we optimize the model's reasoning path selection through the PPO algorithm.We conduct comprehensive experiments on four arithmetic reasoning datasets, demonstrating that our method achieves significant performance improvements compared to state-of-the-art competitors.1

Generating images with both photorealism and multiview 3D consistency is crucial for 3D-aware GANs, yet existing methods struggle to achieve them simultaneously. Improving the photorealism via CNN-based 2D super-resolution can break the strict 3D consistency, while keeping the 3D consistency by learning high-resolution 3D representations for direct rendering often compromises image quality. In this paper, we propose a novel learning strategy, namely 3D-to-2D imitation, which enables a 3D-aware GAN to generate high-quality images while maintaining their strict 3D consistency, by letting the images synthesized by the generator's 3D rendering branch mimic those generated by its 2D super-resolution branch. We also introduce 3D-aware convolutions into the generator for better 3D representation learning, which further improves the image generation quality. With the above strategies, our method reaches FID scores of 5.4 and 4.3 on FFHQ and AFHQ-v2 Cats, respectively, at 512×512 resolution, largely outperforming existing 3D-aware GANs using direct 3D rendering and coming very close to the previous state-of-the-art method that leverages 2D super-resolution. Project website: https://seanchenxy.github.io/Mimic3DWeb.

The SOTA face swap models still suffer the problem of either target identity (i.e., shape) being leaked or the target non-identity attributes (i.e., background, hair) failing to be fully preserved in the final results. We show that this insufficient disentanglement is caused by two flawed designs that were commonly adopted in prior models: (1) counting on only one compressed encoder to represent both the semantic-level non-identity facial attributes(i.e., pose) and the pixel-level non-facial region details, which is contradictory to satisfy at the same time; (2) highly relying on long skip-connections [50] between the encoder and the final generator, leaking a certain amount of target face identity into the result. To fix them, we introduce a new face swap framework called "WSC-swap" that gets rid of skip connections and uses two target encoders to respectively capture the pixel-level non-facial region attributes and the semantic non-identity attributes in the face region. To further reinforce the disentanglement learning for the target encoder, we employ both identity removal loss via adversarial training (i.e., GAN [18]) and the non-identity preservation loss via prior 3DMM models like [11]. Extensive experiments on both FaceForensics++ and CelebA-HQ show that our results significantly outperform previous works on a rich set of metrics, including one novel metric for measuring identity consistency that was completely neglected before.

We introduce a novel framework for one-shot audio-driven talking head generation. Unlike prior works that require additional driving sources for controlled synthesis in a deterministic manner, we instead sample all holistic lip-irrelevant facial motions (i.e. pose, expression, blink, gaze, etc.) to semantically match the input audio while still maintaining both the photo-realism of audio-lip synchronization and overall naturalness. This is achieved by our newly proposed audio-to-visual diffusion prior trained on top of the mapping between audio and non-lip representations. Thanks to the probabilistic nature of the diffusion prior, one big advantage of our framework is it can synthesize diverse facial motion sequences given the same audio clip, which is quite user-friendly for many real applications. Through comprehensive evaluations of public benchmarks, we conclude that (1) our diffusion prior outperforms auto-regressive prior significantly on all the concerned metrics; (2) our overall system is competitive with prior works in terms of audio-lip synchronization but can effectively sample rich and natural-looking lip-irrelevant facial motions while still semantically harmonized with the audio input.

Despite the impressive progress of general face detection, the tuning of hyper-parameters and architectures is still critical for the performance of a domain-specific face detector. Though existing AutoML works can speedup such process, they either require tuning from scratch for a new scenario or do not consider data privacy. To scale up, we derive a new AutoML setting from a platform perspective. In such setting, new datasets sequentially arrive at the platform, where an architecture and hyper-parameter configuration is recommended to train the optimal face detector for each dataset. This, however, brings two major challenges: (1) how to predict the best configuration for any given dataset without touching their raw images due to the privacy concern? and (2) how to continuously improve the AutoML algorithm from previous tasks and offer a better warm-up for future ones? We introduce “HyperFD”, a new privacy-preserving online AutoML framework for face detection. At its core part, a novel meta-feature representation of a dataset as well as its learning paradigm is proposed. Thanks to HyperFD, each local task (client) is able to effectively leverage the learning “experience” of previous tasks without uploading raw images to the platform; meanwhile, the meta-feature extractor is continuously learned to better trade off the bias and variance. Extensive experiments demonstrate the effectiveness and efficiency of our design.

3D single object tracking (3D SOT) in LiDAR point clouds plays a crucial role in autonomous driving. Current approaches all follow the Siamese paradigm based on appearance matching. However, LiDAR point clouds are usually textureless and incomplete, which hinders effective appearance matching. Besides, previous methods greatly overlook the critical motion clues among targets. In this work, beyond 3D Siamese tracking, we introduce a motion-centric paradigm to handle 3D SOT from a new perspective. Following this paradigm, we propose a matching-free two-stage tracker M2-Track. At the 1st-stage, M2-Track localizes the target within successive frames via motion transformation. Then it refines the target box through motion-assisted shape completion at the 2nd-stage. Extensive experiments confirm that M2-Track significantly outperforms previous state-of-the-arts on three large-scale datasets while running at 57FPS (~ 8%, ~ 17% and ~ 22% precision gains on KITTI, NuScenes, and Waymo Open Dataset respectively). Further analysis verifies each component's effectiveness and shows the motioncentric paradigm's promising potential when combined with appearance matching. Code will be made available at https://github.com/Ghostish/Open3DSOT.

Due to the rising concern of data privacy, it's reasonable to assume the local client data can't be transferred to a centralized server, nor their associated identity label is provided. To support continuous learning and fill the last-mile quality gap, we introduce a new problem setup called “local-adaptive face recognition (LaFR)”. Leveraging the environment-specific local data after the deployment of the initial global model, LaFR aims at getting optimal performance by training local-adapted models automatically and un-supervisely, as opposed to fixing their initial global model. We achieve this by a newly proposed embedding cluster model based on Graph Convolution Network (GCN), which is trained via meta-optimization procedure. Compared with previous works, our meta-clustering model can generalize well in unseen local environments. With the pseudo identity labels from the clustering results, we further introduce novel regularization techniques to improve the model adaptation performance. Extensive experiments on racial and internal sensor adaptation demonstrate that our proposed solution is more effective for adapting face recognition models in each specific environment. Meanwhile, we show that LaFR can further improve the global model by a simple federated aggregation over the updated local models.

Face detection is a fundamental problem for many down-stream face applications, and there is a rising demand for faster, more accurate yet support for higher resolution face detectors. Recent smartphones can record a video in 8K resolution, but many of the existing face detectors still fail due to the anchor size and training data. We analyze the failure cases and observe a large number of correct predicted boxes with incorrect confidences. To calibrate these confidences, we propose a confidence ranking network with a pairwise ranking loss to rerank the predicted confidences locally within the same image. Our confidence ranker is model-agnostic, so we can augment the data by choosing the pairs from multiple face detectors during the training, and generalize to a wide range of face detectors during the testing. On WiderFace, we achieve the highest AP on the single-scale, and our AP is competitive with the previous multi-scale methods while being significantly faster. On 8K resolution, our method solves the GPU memory issue and allows us to indirectly train on 8K. We collect 8K resolution test set to show the improvement, and we will release our test set as a new benchmark for future research.

The key challenge for 3D face shape reconstruction is to build the correct dense face correspondence between the deformable mesh and the single input image. Given the ill-posed nature, previous works heavily rely on prior knowledge (such as 3DMM [2]) to reduce depth ambiguity. Although impressive result has been made recently [42, 14, 8], there is still a large room to improve the correspondence so that projected face shape better aligns with the silhouette of each face region (i.e, eye, mouth, nose, cheek, etc.) on the image. To further reduce the ambiguities, we present a novel framework called “Reinforced Differentiable Attributes” (“ReDA”) which is more general and effective than previous Differentiable Rendering (“DR”). Specifically, we first extend from color to more broad attributes, including the depth and the face parsing mask. Secondly, unlike the previous Z-buffer rendering, we make the rendering to be more differentiable through a set of convolution operations with multi-scale kernel sizes. In the meanwhile, to make “ReDA” to be more successful for 3D face recon-struction, we further introduce a new free-form deformation layer that sits on top of 3DMM to enjoy both the prior knowledge and out-of-space modeling. Both techniques can be easily integrated into existing 3D face reconstruction pipeline. Extensive experiments on both RGB and RGB-D datasets show that our approach outperforms prior arts.

Previous methods for audio-driven talking head generation assume the input audio to be clean with a neutral tone. As we show empirically, one can easily break these systems by simply adding certain background noise to the utterance or changing its emotional tone (to for example, sad). To make talking head generation robust to such variations, we propose an explicit audio representation learning framework that disentangles audio sequences into various factors such as phonetic content, emotional tone, background noise and others. We conduct experiments to validate that when conditioned on disentangled content representation, the generated mouth movement by our model is significantly more accurate than previous approaches (without disentangled learning) in the presence of noise and emotional variations. We further demonstrate that our framework is compatible with current state-of-the-art approaches by replacing their original component to learn audio based representation with ours. To the best of our knowledge, this is the first work which improves the performance of talking head generation through a disentangled audio representation perspective, which is important for many real-world applications.

We present an automatic moment capture system that runs in real-time on mobile cameras. The system is designed to run in the viewfinder mode and capture a burst sequence of frames before and after the shutter is pressed. For each frame, the system predicts in real-time a goodness score, based on which the best moment in the burst can be selected immediately after the shutter is released. We develop a highly efficient deep neural network ranking model, which implicitly learns a latent relative attribute space to capture subtle visual differences within a sequence of burst images. The overall goodness is computed as a linear aggregation of the goodnesses of all the latent attributes. To obtain a compact model which can run on mobile devices in real-time, we have explored and evaluated a wide range of network design choices, taking into account the constraints of model size, computational cost, and accuracy. Extensive studies show that the best frame predicted by our model hit users' top-1 (out of 11 on average) choice for 64.1% cases and top-3 choices for 86.2% cases. Moreover, the model (only 0.47M Bytes) can run in real time on mobile devices, e.g. 13ms on iPhone 7.

Facial motion retargeting is an important problem in both computer graphics and vision, which involves capturing the performance of a human face and transferring it to another 3D character. Learning 3D morphable model (3DMM) parameters from 2D face images using convolutional neural networks is common in 2D face alignment, 3D face reconstruction etc. However, existing methods either require an additional face detection step before retargeting or use a cascade of separate networks to perform detection followed by retargeting in a sequence. In this paper, we present a single end-to-end network to jointly predict the bounding box locations and 3DMM parameters for multiple faces. First, we design a novel multitask learning framework that learns a disentangled representation of 3DMM parameters for a single face. Then, we leverage the trained single face model to generate ground truth 3DMM parameters for multiple faces to train another network that performs joint face detection and motion retargeting for images with multiple faces. Experimental results show that our joint detection and retargeting network has high face detection accuracy and is robust to extreme expressions and poses while being faster than state-of-the-art methods.

We propose a reinforcement learning approach for real-time exposure control of a mobile camera that is personalizable. Our approach is based on Markov Decision Process (MDP). In the camera viewfinder or live preview mode, given the current frame, our system predicts the change in exposure so as to optimize the trade-off among image quality, fast convergence, and minimal temporal oscillation. We model the exposure prediction function as a fully convolutional neural network that can be trained through Gaussian policy gradient in an end-to-end fashion. As a result, our system can associate scene semantics with exposure values; it can also be extended to personalize the exposure adjustments for a user and device. We improve the learning performance by incorporating an adaptive metering module that links semantics with exposure. This adaptive metering module generalizes the conventional spot or matrix metering techniques. We validate our system using the MIT FiveK [1] and our own datasets captured using iPhone 7 and Google Pixel. Experimental results show that our system exhibits stable real-time behavior while improving visual quality compared to what is achieved through native camera control.

Retouching can significantly elevate the visual appeal of photos, but many casual photographers lack the expertise to do this well. To address this problem, previous works have proposed automatic retouching systems based on supervised learning from paired training images acquired before and after manual editing. As it is difficult for users to acquire paired images that reflect their retouching preferences, we present in this article a deep learning approach that is instead trained on unpaired data, namely, a set of photographs that exhibits a retouching style the user likes, which is much easier to collect. Our system is formulated using deep convolutional neural networks that learn to apply different retouching operations on an input image. Network training with respect to various types of edits is enabled by modeling these retouching operations in a unified manner as resolution-independent differentiable filters. To apply the filters in a proper sequence and with suitable parameters, we employ a deep reinforcement learning approach that learns to make decisions on what action to take next, given the current state of the image. In contrast to many deep learning systems, ours provides users with an understandable solution in the form of conventional retouching edits rather than just a “black-box” result. Through quantitative comparisons and user studies, we show that this technique generates retouching results consistent with the provided photo set.

Improvements in color constancy have arisen from the use of convolutional neural networks (CNNs). However, the patch-based CNNs that exist for this problem are faced with the issue of estimation ambiguity, where a patch may contain insufficient information to establish a unique or even a limited possible range of illumination colors. Image patches with estimation ambiguity not only appear with great frequency in photographs, but also significantly degrade the quality of network training and inference. To overcome this problem, we present a fully convolutional network architecture in which patches throughout an image can carry different confidence weights according to the value they provide for color constancy estimation. These confidence weights are learned and applied within a novel pooling layer where the local estimates are merged into a global solution. With this formulation, the network is able to determine what to learn and how to pool automatically from color constancy datasets without additional supervision. The proposed network also allows for end-to-end training, and achieves higher efficiency and accuracy. On standard benchmarks, our network outperforms the previous state-of-the-art while achieving 120× greater efficiency.

Cinemagraphs are a compelling way to convey dynamic aspects of a scene. In these media, dynamic and still elements are juxtaposed to create an artistic and narrative experience. Creating a high-quality, aesthetically pleasing cinemagraph requires isolating objects in a semantically meaningful way and then selecting good start times and looping periods for those objects to minimize visual artifacts (such a tearing). To achieve this, we present a new technique that uses object recognition and semantic segmentation as part of an optimization method to automatically create cinemagraphs from videos that are both visually appealing and semantically meaningful. Given a scene with multiple objects, there are many cinemagraphs one could create. Our method evaluates these multiple candidates and presents the best one, as determined by a model trained to predict human preferences in a collaborative way. We demonstrate the effectiveness of our approach with multiple results and a user study.

The iterations of many sparse estimation algorithms are comprised of a fixed linear filter cascaded with a thresholding nonlinearity, which collectively resemble a typical neural network layer. Consequently, a lengthy sequence of algorithm iterations can be viewed as a deep network with shared, hand-crafted layer weights. It is therefore quite natural to examine the degree to which a learned network model might act as a viable surrogate for traditional sparse estimation in domains where ample training data is available. While the possibility of a reduced computational budget is readily apparent when a ceiling is imposed on the number of layers, our work primarily focuses on estimation accuracy. In particular, it is well-known that when a signal dictionary has coherent columns, as quantified by a large RIP constant, then most tractable iterative algorithms are unable to find maximally sparse representations. In contrast, we demonstrate both theoretically and empirically the potential for a trained deep network to recover minimal $\ell_0$-norm representations in regimes where existing methods fail. The resulting system, which can effectively learn novel iterative sparse estimation algorithms, is deployed on a practical photometric stereo estimation problem, where the goal is to remove sparse outliers that can disrupt the estimation of surface normals from a 3D scene.

We address the problem of traffic sign recognition in a light detection and ranging (LIDAR)-equipped vehicle. With the help of 3-D LIDAR points, the 2-D multiview sign images will be easily detected from the captured images of street signs. After detection, the sign recognition problem is formulated as a multiview object recognition task. We develop a metric-learning-based template matching approach for this task and learn a distance metric between the captured images and the corresponding sign templates. For each sign, recognition is done via soft voting by the recognition results of its corresponding multiview images. We propose a latent structural support vector machine (SVM)-based weakly supervised metric learning (WSMLR) method to learn the metric and a reliability classifier. The reliability classifier is used to determine each image's reliability, which serves as each image's weight in both the learning and soft voting procedure. We evaluate the proposed method for multiview traffic sign recognition on a multiview traffic sign data set with 112 categories and observe very encouraging results compared with other state-of-the-art methods. In addition, the method can be customized to solve the single-view sign recognition. The performance of our method for single-view sign recognition is tested on two public data sets, showing that our method is comparable with other competitive ones.

Photo retouching enables photographers to invoke dramatic visual impressions by artistically enhancing their photos through stylistic color and tone adjustments. However, it is also a time-consuming and challenging task that requires advanced skills beyond the abilities of casual photographers. Using an automated algorithm is an appealing alternative to manual work, but such an algorithm faces many hurdles. Many photographic styles rely on subtle adjustments that depend on the image content and even its semantics. Further, these adjustments are often spatially varying. Existing automatic algorithms are still limited and cover only a subset of these challenges. Recently, deep learning has shown unique abilities to address hard problems. This motivated us to explore the use of deep neural networks (DNNs) in the context of photo editing. In this article, we formulate automatic photo adjustment in a manner suitable for this approach. We also introduce an image descriptor accounting for the local semantics of an image. Our experiments demonstrate that training DNNs using these descriptors successfully capture sophisticated photographic styles. In particular and unlike previous techniques, it can model local adjustments that depend on image semantics. We show that this yields results that are qualitatively and quantitatively better than previous work.

We present a new approach for feature pooling in human action recognition. Instead of partitioning videos at predefined uniform intervals in a spatial-temporal volume as done with spatial pyramid matching, our method adaptively partitions in a pooling attribute space, defined by multiple trajectory-based cues. The pooling attributes include individual spatial and temporal coordinates of a trajectory, as well as its motion saliency, curvature, and scale. To determine partitions of the attribute space in an adaptive manner, we utilize KD-trees that separate trajectories based on their distributions within the attribute space. The generated pooling volumes are jointly utilized for action recognition via SVM weights learned by Multiple Kernel Learning. Through extensive experimentation on major benchmarks, it is shown that this adaptive pooling over multiple trajectory attributes leads to significant improvements in recognition performance.

Recent work on scene classification still makes use of generic CNN features in a rudimentary manner. In this paper, we present a novel pipeline built upon deep CNN features to harvest discriminative visual objects and parts for scene classification. We first use a region proposal technique to generate a set of high-quality patches potentially containing objects, and apply a pre-trained CNN to extract generic deep features from these patches. Then we perform both unsupervised and weakly supervised learning to screen these patches and discover discriminative ones representing category-specific objects and parts. We further apply discriminative clustering enhanced with local CNN fine-tuning to aggregate similar objects and parts into groups, called meta objects. A scene image representation is constructed by pooling the feature response maps of all the learned meta objects at multiple spatial scales. We have confirmed that the scene image representation obtained using this new pipeline is capable of delivering state-of-the-art performance on two popular scene benchmark datasets, MIT Indoor 67 [22] and Sun397 [31].

With the growing popularity of short-form video sharing platforms such as Instagram and Vine, there has been an increasing need for techniques that automatically extract highlights from video. Whereas prior works have approached this problem with heuristic rules or supervised learning, we present an unsupervised learning approach that takes advantage of the abundance of user-edited videos on social media websites such as YouTube. Based on the idea that the most significant sub-events within a video class are commonly present among edited videos while less interesting ones appear less frequently, we identify the significant sub-events via a robust recurrent auto-encoder trained on a collection of user-edited videos queried for each particular class of interest. The auto-encoder is trained using a proposed shrinking exponential loss function that makes it robust to noise in the web-crawled training data, and is configured with bidirectional long short term memory (LSTM) [5] cells to better model the temporal structure of highlight segments. Different from supervised techniques, our method can infer highlights using only a set of downloaded edited videos, without also needing their pre-edited counterparts which are rarely available online. Extensive experiments indicate the promise of our proposed solution in this challenging unsupervised setting.

With the improved accessibility to an exploding amount of video data and growing demands in a wide range of video analysis applications, video-based action recognition/classification becomes an increasingly important task in computer vision. In this paper, we propose a two-layer structure for action recognition to automatically exploit a mid-level ``acton'' representation. The actons are learned via a new max-margin multi-channel multiple instance learning framework. The learned actons (with no requirement for detailed manual annotations) thus observe a property of being compact, informative, discriminative, and easy to scale. This is different from the standard unsupervised (e.g. k-means) or supervised (e.g. random forests) coding strategies in action recognition. Applying the learned actons in our two-layer structure yields the state-of-the-art classification performance on Youtube and HMDB51 datasets.

Dictionary learning has became an increasingly important task in machine learning, as it is fundamental to the representation problem. A number of emerging techniques specifically include a codebook learning step, in which a critical knowledge abstraction process is carried out. Existing approaches in dictionary (codebook) learning are either generative (unsupervised e.g. k-means) or discriminative (supervised e.g. extremely randomized forests). In this paper, we propose a multiple instance learning (MIL) strategy (along the line of weakly supervised learning) for dictionary learning. Each code is represented by a classifier, such as a linear SVM, which naturally performs metric fusion for multi-channel features. We design a formulation to simultaneously learn mixtures of codes by maximizing classification margins in MIL. State-of-the-art results are observed in image classification benchmarks based on the learned codebooks, which observe both compactness and effectiveness.

Color and tone adjustments are among the most frequent image enhancement operations. We define a color and tone style as a set of explicit or implicit rules governing color and tone adjustments. Our goal in this paper is to learn implicit color and tone adjustment rules from examples. That is, given a set of examples, each of which is a pair of corresponding images before and after adjustments, we would like to discover the underlying mathematical relationships optimally connecting the color and tone of corresponding pixels in all image pairs. We formally define tone and color adjustment rules as mappings, and propose to approximate complicated spatially varying nonlinear mappings in a piecewise manner. The reason behind this is that a very complicated mapping can still be locally approximated with a low-order polynomial model. Parameters within such low-order models are trained using data extracted from example image pairs. We successfully apply our framework in two scenarios, low-quality photo enhancement by transferring the style of a high-end camera, and photo enhancement using styles learned from photographers and designers.

It is often important for designers and photographers to convey or enhance desired color themes in their work. A color theme is typically defined as a template of colors and an associated verbal description. This paper presents a data-driven method for enhancing a desired color theme in an image. We formulate our goal as a unified optimization that simultaneously considers a desired color theme, texture-color relationships as well as automatic or user-specified color constraints. Quantifying the difference between an image and a color theme is made possible by color mood spaces and a generalization of an additivity relationship for two-color combinations. We incorporate prior knowledge, such as texture-color relationships, extracted from a database of photographs to maintain a natural look of the edited images. Experiments and a user study have confirmed the effectiveness of our method.