Rewriting inputs to improve frozen downstream models has become a common strategy in modern NLP pipelines. Prior work on incremental dialogue discourse parsing (DDP) shows that supervised clarification models can rewrite fragmentary or underspecified utterances—such as resolving ellipsis or references—to improve parsing accuracy. In this work, we revisit this idea under realistic deployment conditions, where no clarification supervision is available and the clarifier must rely on zero-shot prompting or feedback from a frozen parser. Across three Segmented Discourse Representation Theory (SDRT) datasets and multiple parsers, we find that last-utterance clarification is far less reliable than suggested by supervised settings. Parser-agnostic rewriting often introduces more regressions than repairs, as edits that enable fixes also disrupt discourse cues relied upon by the parser. A best-of-8 rewriting analysis further reveals a practical ceiling: a large fraction of errors are not repairable through input rewriting alone. A parser-aware clarifier trained with GRPO reduces regressions by up to 37% by learning conservative abstention, yet still fails to produce selectivity-aware clarifications that consistently improve parsing. Together, these findings recast clarification as a selective intervention problem. We identify rewritability prediction—deciding whether an utterance is repairable before intervention—as the key missing capability for input-side optimization of frozen discourse parsers, and a critical direction for improving agentic pipelines more broadly.[Data and code are available at https://github.com/ounlp/Clarification-for-DDP.]

Learning sign language is not only a gateway to linguistic competence but also to cultural participation, self-expression, and meaningful interaction within the Deaf community. Recent advances in sign language education technologies have made notable progress in supporting vocabulary acquisition and sentence-level translation. However, dialogue, as one critical component of natural language use, remains largely absent from existing systems. We identify two underlying challenges that contribute to this gap: the lack of recognition robustness to viewpoint variation, which constrains expressive freedom during signing; and the limited semantic modeling capabilities needed to support discourse-level interpretation and interaction. To address these issues, we propose a learning-centered system that integrates pose-based multi-view augmentation and a multi-agent language modeling workflow. This system supports free-form input across diverse signing perspectives and provides structured feedback across word, sentence, and dialogue levels. Built as a modular and deployable platform, the system demonstrates strong recognition performance and learner engagement across varied input conditions. Through this integration of spatial robustness and semantic scaffolding, our work advances the design of sign language learning technologies toward more interactive, expressive, and pedagogically grounded experiences. However, the current system remains limited by its focus on successive (non-continuous) signing and by its moderate vocabulary and restricted pedagogical scope. Future work will therefore extend lexical coverage, incorporate more natural continuous signing, and investigate richer educational functions. • A View-invariant Pose-based Isolated Auslan Sign Recognition. • A Multi-agent Workflow Providing Transparent Dialogue Support. • An Interactive, Learner-paced Design for Auslan Learning System.

Abstract Like spoken languages, a single sign language expression could correspond to multiple valid textual interpretations. Hence, learning a rigid one-to-one mapping for sign language translation (SLT) models might be inadequate, particularly in the case of limited data. In this work, we introduce a Diverse Sign Language Translation (DivSLT) task, aiming to generate diverse yet accurate translations for sign language videos. Firstly, we employ large language models (LLM) to generate multiple references for the widely-used CSL-Daily and PHOENIX14T SLT datasets. Here, native speakers are only invited to touch up inaccurate references, thus significantly improving the annotation efficiency. Secondly, we provide a benchmark model to spur research in this task. Specifically, we investigate multi-reference training strategies enabling our DivSLT model to achieve diverse translations. Then, to enhance translation accuracy, we employ the max-reward-driven reinforcement learning objective that maximizes the reward of the translated result. Additionally, we utilize multiple metrics to assess the accuracy, diversity, and semantic precision of the DivSLT task. Experimental results on the enriched datasets demonstrate that our DivSLT method achieves not only better translation performance but also diverse translation results.

• Introduces CAPEL: ensembles multiple prompts in logit space for VLM adaptation. • Preserves multimodal class structure and supports fully batched inference. • Uses cluster-preserving entropy to stabilize diverse prompt subclassifiers. • Learns prompt-level attention and enables simple post-hoc pruning of heads. • Demonstrates generalization across backbones and tasks, including robustness and segmentation. Vision-language models (VLMs) such as CLIP achieve zero-shot transfer across various tasks by pre-training on numerous image-text pairs. These models often benefit from using an ensemble of context prompts to represent a class. Despite being effective, conventional prompt ensembling that averages textual features of context prompts often yields suboptimal results. This is because feature averaging shifts the class centroids away from the true class distribution. To address this issue, we propose the Cluster-Aware Prompt Ensemble Learning (CAPEL) framework, which preserves the cluster nature of context prompts. CAPEL classifies images into one of several class clusters, each represented by a distinct prompt. Instead of ensembling prompts in the feature space, we perform ensembling in the classification logits space, aligning better with the visual feature distribution. To further optimize prompt fine-tuning while maintaining cluster-specific discriminative power, we introduce a cluster-preserving regularization term. This ensures that prompts remain distinct and specialized for different clusters, preventing collapse into a uniform direction. Additionally, we integrate an adaptive prompt weighting technique to dynamically adjust the attention weights for flawed or ambiguous prompts, ensuring robust performance across diverse datasets and tasks.

Large-scale text-to-image (T2I) diffusion models have revolutionized image generation, enabling the synthesis of highly detailed visuals from textual descriptions. However, these models may inadvertently generate inappropriate content, such as copyrighted works or offensive images. While existing methods attempt to eliminate specific unwanted concepts, they often fail to ensure robust removal-allowing the concept to reappear in subtle forms. For instance, a model may successfully avoid generating images in Van Gogh's style when explicitly prompted with “Van Gogh”, yet still reproduce his signature artwork when given the prompt “Starry Night”. In this paper, we propose SAFER, a novel and efficient approach for thoroughly removing target concepts from diffusion models. At a high level, SAFER is inspired by the observed low-dimensional structure of the text embedding space. The method first identifies a concept-specific subspace$\mathcal {S}_{c}$associated with the target concept$c$. It then projects the prompt embeddings onto the complementary subspace of$\mathcal {S}_{c}$, effectively erasing the concept from the generated images. Since concepts can be abstract and difficult to fully capture using natural language alone, we employ textual inversion to learn an optimized embedding of the target concept from a reference image. This enables more precise subspace estimation and enhances removal performance. Furthermore, we introduce a subspace expansion strategy to ensure comprehensive and robust concept erasure. Extensive experiments demonstrate that SAFER consistently and effectively erases unwanted concepts from diffusion models while preserving generation quality.

Current compressed video super-resolution methods have achieved promising performance, but they often assume that an input video is compressed under low-delay configurations. However, under random access configurations, those methods might struggle to leverage the metadata effectively due to the large variations of metadata in different compression configurations. In this work, we propose a Compression-Oriented Video Super-Resolution (COVSR) method that can address video super-resolution for both low-delay and random-access configurations. Specifically, we first introduce an efficient compression-aware propagation (ECAP) module that dynamically adjusts propagation routes in accordance with the compression configurations. Since existing methods require reconstructing frames in a frame-by-frame manner, it is difficult to achieve efficient parallelization. However, we find that by slightly relaxing sequential dependencies, our ECAP can significantly improve inference speed. Furthermore, existing methods typically perform alignment between adjacent frames or adjacent features. However, since ECAP may propagate features along non-adjacent reference routes, it introduces new challenges for accurate cross-frame feature alignment. In response, we propose a metadata-driven alignment (MDA) module that refines cross-frame motion vectors into dense, feature-level flow offsets, enabling precise alignment across temporally distant features. Extensive experimental results demonstrate that our COVSR not only achieves efficient and superior super-resolution performance but also is generalizable to various compression configurations. Our code will be available at https://covsr.github.io.

Accurately localizing audible objects based on audio-visual cues is the core objective of audio-visual segmentation. Most previous methods emphasize spatial or temporal multi-modal modeling, yet overlook challenges from ambiguous audio-visual correspondences—such as nearby visually similar but acoustically different objects and frequent shifts in objects’ sounding status. Consequently, they may struggle to reliably correlate audio and visual cues, leading to over-or under-segmentation. To address these limitations, we propose a novel framework with two primary components: an audio-guided modality alignment (AMA) module and an uncertainty estimation (UE) module. Instead of indiscriminately correlating audio-visual cues through a global attention mechanism, AMA performs audio-visual interactions within multiple groups and consolidates group features into compact representations based on their responsiveness to audio cues, effectively directing the model’s attention to audio-relevant areas. Leveraging contrastive learning, AMA further distinguishes sounding regions from silent areas by treating features with strong audio responses as positive samples and weaker responses as negatives. Additionally, UE integrates spatial and temporal information to identify high-uncertainty regions caused by frequent changes in sound state, reducing prediction errors by lowering confidence in these areas. Experimental results demonstrate that our approach achieves superior accuracy compared to existing state-of-the-art methods, particularly in challenging scenarios where traditional approaches struggle to maintain reliable segmentation.

Sound-guided object segmentation has drawn considerable attention for its potential to enhance multimodal perception. Previous methods primarily focus on developing advanced architectures to facilitate effective audio-visual interactions, without fully addressing the inherent challenges posed by audio natures, i.e., (1) feature confusion due to the overlapping nature of audio signals, and (2) audiovisual matching difficulty from the varied sounds produced by the same object. To address these challenges, we propose Dynamic Derivation and Elimination (DDESeg): a novel audio-visual segmentation framework. Specifically, to mitigate feature confusion, DDESeg reconstructs the semantic content of the mixed audio signal by enriching the distinct semantic information of each individual source, deriving representations that preserve the unique characteristics of each sound. To reduce the matching difficulty, we introduce a discriminative feature learning module, which enhances the semantic distinctiveness of generated audio representations. Considering that not all derived audio representations directly correspond to visual features (e.g., off-screen sounds), we propose a dynamic elimination module to filter out non-matching elements. This module facilitates targeted interaction between sounding regions and relevant audio semantics. By scoring the interacted features, we identify and filter out irrelevant audio information, ensuring accurate audio-visual alignment. Comprehensive experiments demonstrate that our framework achieves superior performance in AVS datasets. Our code is here.

Bitstream-corrupted video recovery aims to fill in realistic video content due to bitstream corruption during video storage or transmission. Most existing methods typically assume that the predefined masks of the corrupted regions are known in advance. However, manually annotating these masks is laborious and time-consuming, limiting the applicability of existing methods in real-world scenarios. Therefore, we expect to relax this assumption by defining a new blind video recovery setting where the recovery of corrupted regions does not rely on predefined masks. There are two significant challenges in this setting: (i) without predefined masks, how accurately can a model identify the regions requiring recovery? (ii) how to recover contents from extensive and irregular regions, especially when large portions of frames are severely degraded? To address these challenges, we introduce a Metadata-Guided Diffusion Model, dubbed M-GDM. To enable a diffusion model focusing on the corrupted regions, we leverage intrinsic video metadata as a corruption indicator and design a dual-stream metadata encoder. This encoder first embeds the motion vectors and frame types of a video separately and then merges them into a unified metadata representation. The metadata representation will interact with the corrupted latent feature through cross-attention mechanisms at each diffusion step. Meanwhile, to preserve the intact regions, we propose a prior-driven mask predictor that generates pseudo masks for the corrupted regions by leveraging the metadata prior and diffusion prior. These pseudo masks enable the separation and recombination of intact and recovered regions through hard masking. However, imperfections in pseudo mask predictions and hard masking processes often result in boundary artifacts. Thus, we introduce a post-refinement module that refines the hard-masked outputs, enhancing the consistency between intact and recovered regions. Extensive experiment results validate the effectiveness of our method and demonstrate its superiority in the blind video recovery task.

Character customization, or ’face crafting,’ is a vital feature in role-playing games (RPGs), enhancing player engagement by enabling the creation of personalized avatars. Existing automated methods often struggle with generalizability across diverse game engines due to their reliance on the intermediate constraints of specific image domain and typically support only one type of input, either text or image. To overcome these challenges, we introduce EasyCraft, an innovative end-to-end feedforward framework that automates character crafting by uniquely supporting both text and image inputs. Our approach employs a translator capable of converting facial images of any style into crafting parameters. We first establish a unified feature distribution in the translator’s image encoder through self-supervised learning on a large-scale dataset, enabling photos of any style to be embedded into a unified feature representation.Subsequently, we map this unified feature distribution to crafting parameters specific to a game engine, a process that can be easily adapted to most game engines and thus enhances EasyCraft’s generalizability. By integrating text-to-image techniques with our translator, EasyCraft also facilitates precise, text-based character crafting. EasyCraft’s ability to integrate diverse inputs significantly enhances the versatility and accuracy of avatar creation. Extensive experiments on two RPG games demonstrate the effectiveness of our method, achieving state-of-the-art results and facilitating adaptability across various avatar engines.

Human pose estimation is a critical task in computer vision for applications in sports analysis, healthcare monitoring, and human-computer interaction. However, existing human pose datasets are collected either from custom-configured laboratories with complex devices or they only include data on single individuals, and both types typically capture daily activities. In this paper, we introduce the M3GYM dataset, a large-scale multimodal, multi-view, and multi-person pose dataset collected from a real gym to address the limitations of existing datasets. Specifically, we collect videos for 82 sessions from the gym, each session lasting between 40 to 60 minutes. These videos are gathered by 8 cameras, including over 50 subjects and 47 million frames. These sessions include 51 Normal fitness exercise sessions as well as 17 Pilates and 14 Yoga sessions. The exercises cover a wide range of poses and typical fitness activities, particularly in Yoga and Pilates, featuring poses with stretches, bends, and twists, e.g., humble warrior, fire hydrants and knee hover side twists. Each session involves multiple subjects, leading to significant self-occlusion and mutual occlusion in single views. Moreover, the gym has two symmetric floor mirrors, a feature not seen in previous datasets, and seven lighting conditions. We provide frame-level multimodal annotations, including 2D&3D keypoints, subject IDs, and meshes. Additionally, M3GYM uniquely offers labels for over 500 actions along with corresponding assessments from sports experts. We benchmark a variety of state-of-the-art methods for several tasks, i.e., 2D human pose estimation, single-view and multi-view 3D human pose estimation, and human mesh recovery. To simulate real-world applications, we also conduct cross-domain experiments across Normal, Yoga, and Pilates sessions. The results show that M3GYM significantly improves model generalization in complex real-world settings. The project is available here.

Existing deep learning methods have made significant progress in gait recognition. Quantization can facilitate the application of gait models as a model-agnostic general compression technique. Typically, appearance-based models binarize inputs into silhouette sequences. However, mainstream quantization methods prioritize minimizing task loss over quantization error, which is detrimental to gait recognition with binarized inputs. To address this, we propose a differentiable soft quantizer, which better simulates the gradient of the round function during backpropagation. This enables the network to learn from subtle input perturbations. However, our theoretical analysis and empirical studies reveal that directly applying the soft quantizer can hinder network convergence. We addressed this issue by adopting a two-stage training strategy, introducing a soft quantizer during the fine-tuning phase. However, in the first stage of training, we observed a significant change in the output distribution of different samples in the feature space compared to the full-precision network. It is this change that led to a loss in performance. Based on this, we propose an Inter-class Distance-guided Calibration (IDC) strategy to preserve the relative distance between the embeddings of samples with different labels. Extensive experiments validate the effectiveness of our approach, demonstrating state-of-the-art accuracy across various settings and datasets. The code will be made publicly available.

3D cars are commonly used in self-driving systems, virtual/augmented reality, and games. However, existing 3D car datasets are either synthetic or low-quality, limiting their applications in practical scenarios and presenting a significant gap toward high-quality real-world 3D car datasets. In this paper, we propose the first large-scale 3D real car dataset, termed 3DRealCar, offering three distinctive features. (1) \textbf{High-Volume}: 2,500 cars are meticulously scanned by smartphones, obtaining car images and point clouds with real-world dimensions; (2) \textbf{High-Quality}: Each car is captured in an average of 200 dense, high-resolution 360-degree RGB-D views, enabling high-fidelity 3D reconstruction; (3) \textbf{High-Diversity}: The dataset contains various cars from over 100 brands, collected under three distinct lighting conditions, including reflective, standard, and dark. Additionally, we offer detailed car parsing maps for each instance to promote research in car parsing tasks. Moreover, we remove background point clouds and standardize the car orientation to a unified axis for the reconstruction only on cars and controllable rendering without background. We benchmark 3D reconstruction results with state-of-the-art methods across different lighting conditions in 3DRealCar. Extensive experiments demonstrate that the standard lighting condition part of 3DRealCar can be used to produce a large number of high-quality 3D cars, improving various 2D and 3D tasks related to cars. Notably, our dataset brings insight into the fact that recent 3D reconstruction methods face challenges in reconstructing high-quality 3D cars under reflective and dark lighting conditions. \textcolor{red}{\href{https://xiaobiaodu.github.io/3drealcar/}{Our dataset is here.}}

Machine unlearning aims to remove the influence of specific samples from a trained model. A key challenge in this process is over-unlearning, where the model's performance on the remaining data significantly drops due to the change in the model's parameters. Existing unlearning algorithms depend on the remaining data to prevent this issue. As such, these methods are inapplicable in a more practical scenario, where only the unlearning samples are available (i.e., zero-shot unlearning). This paper presents a novel framework, ZS-PAG, to fill this gap. Our approach offers three key innovations: (1) we approximate the inaccessible remaining data by generating adversarial samples; (2) leveraging the generated samples, we pinpoint a specific subspace to perform the unlearning process, therefore preventing over-unlearning in the challenging zero-shot scenario; and (3) we consider the influence of the unlearning process on the remaining samples and design an influence-based pseudo-labeling strategy. As a result, our method further improves the model's performance after unlearning. The proposed method holds a theoretical guarantee, and experiments on various benchmarks validate the effectiveness and superiority of our proposed method over several baselines.

Retina images provide a noninvasive view of the central nervous system and microvasculature, making it essential for clinical applications. Changes in the retina often indicate both ophthalmic and systemic diseases, aiding in diagnosis and early intervention. While deep learning algorithms have advanced retina image analysis, a comprehensive review of related datasets, tasks, and benchmarking is still lacking. In this survey, we systematically categorize existing retina image datasets based on their available data modalities, and review the tasks these datasets support in multimodal retina image analysis. We also explain key evaluation metrics used in various retina image analysis benchmarks. By thoroughly examining current datasets and methods, we highlight the challenges and limitations in existing benchmarks and discuss potential research topics in the field. We hope this work will guide future retina analysis methods and promote the shared use of existing data across different tasks.

Existing Moment retrieval (MR) methods focus on Single-Moment Retrieval (SMR). However, one query can correspond to multiple relevant moments in real-world applications. This makes the existing datasets and methods insufficient for video temporal grounding. By revisiting the gap between current MR tasks and real-world applications, we introduce a high-quality datasets called QVHighlights Multi-Moment Dataset (QV-M$^2$), along with new evaluation metrics tailored for multi-moment retrieval (MMR). QV-M$^2$ consists of 2,212 annotations covering 6,384 video segments. Building on existing efforts in MMR, we propose a framework called FlashMMR. Specifically, we propose a Multi-moment Post-verification module to refine the moment boundaries. We introduce constrained temporal adjustment and subsequently leverage a verification module to re-evaluate the candidate segments. Through this sophisticated filtering pipeline, low-confidence proposals are pruned, and robust multi-moment alignment is achieved. We retrain and evaluate 6 existing MR methods on QV-M$^2$ and QVHighlights under both SMR and MMR settings. Results show that QV-M$^2$ serves as an effective benchmark for training and evaluating MMR models, while FlashMMR provides a strong baseline. Specifically, on QV-M$^2$, it achieves improvements over prior SOTA method by 3.00% on G-mAP, 2.70% on mAP@3+tgt, and 2.56% on mR@3. The proposed benchmark and method establish a foundation for advancing research in more realistic and challenging video temporal grounding scenarios. Code is released at https://github.com/Zhuo-Cao/QV-M2.

Text-guided Video Temporal Grounding (VTG) aims to localize relevant segments in untrimmed videos based on textual descriptions, encompassing two subtasks: Moment Retrieval (MR) and Highlight Detection (HD). Although previous typical methods have achieved commendable results, it is still challenging to retrieve short video moments. This is primarily due to the reliance on sparse and limited decoder queries, which significantly constrain the accuracy of predictions. Furthermore, suboptimal outcomes often arise because previous methods rank predictions based on isolated predictions, neglecting the broader video context. To tackle these issues, we introduce FlashVTG, a framework featuring a Temporal Feature Layering (TFL) module and an Adaptive Score Refinement (ASR) module. The TFL module replaces the traditional decoder structure to capture nuanced video content variations across multiple temporal scales, while the ASR module improves prediction ranking by integrating context from adjacent moments and multi-temporal-scale features. Extensive experiments demonstrate that FlashVTG achieves state-of-the-art performance on four widely adopted datasets in both MR and HD. Specifically, on the QVHighlights dataset, it boosts mAP by 5.8% for MR and 3.3% for HD. For short-moment retrieval, FlashVTG increases mAP to 125% of previous SOTA performance. All these improvements are made without adding training burdens, underscoring its effectiveness. Our code is available at https://github.com/Zhuo-Cao/FlashVTG.

Text-Video Retrieval (TVR) methods typically match query-candidate pairs by aligning text and video features in coarse-grained, fine-grained, or combined (coarse-to-fine) manners. However, these frameworks predominantly employ a one(query)-to-one(candidate) alignment paradigm, which struggles to discern nuanced differences among candidates, leading to frequent mismatches. Inspired by Comparative Judgement in human cognitive science, where decisions are made by directly comparing items rather than evaluating them independently, we propose TokenBinder. This innovative two-stage TVR framework introduces a novel one-to-many coarse-to-fine alignment paradigm, imitating the human cognitive process of identifying specific items within a large collection. Our method employs a Focused-view Fusion Network with a sophisticated cross-attention mechanism, dynamically aligning and comparing features across multiple videos to capture finer nuances and contextual variations. Extensive experiments on six benchmark datasets confirm that TokenBinder substantially outperforms existing state-of-the-art methods. These results demonstrate its robustness and the effectiveness of its fine-grained alignment in bridging intra- and inter-modality information gaps in TVR tasks. Code is avaliable at https://github.com/bingqingzhang/TokenBinder.

Misinformation is a significant societal issue with potentially severe consequences. It appears in text, image, audio, and video modalities, encompassing various categories such as unimodal deception (fact-conflicting, AI-generated & offensive content) and cross-modal inconsistencies. However, current detection approaches often focus on text and image, overlooking the growing prevalence of misinformation in audio and video content. Moreover, these methods typically tend to address only one or two types of misinformation, failing to address all categories simultaneously. These detectors are also usually designed to make judgments without providing explanations, reducing transparency and limiting their broader applicability. To address these issues, we propose MDAM3, a Misinformation Detection and Analysis Framework for Multitype Multimodal Media. MDAM3 analyzes each input in internal detection and examines relationships across modalities to identify inconsistencies. It utilizes web resources and integrates Large Vision-Language Models (LVLMs) to deliver accurate detection results along with detailed analysis. To evaluate MDAM3, we curate MDAM3-DB, a specialized multitype multimodal misinformation dataset. A user study is conducted to explore MDAM3's usability, interpretability, and effectiveness. We hope this research contributes to advancing misinformation detection methodologies and provides valuable insights for developing robust multimodal analysis tools.

Audio-driven talking head generation has drawn growing attention. To produce talking head videos with desired facial expressions, previous methods rely on extra reference videos to provide expression information, which may be difficult to find and hence limits their usage. In this work, we propose TalkCLIP, a framework that can generate talking heads where the expressions are specified by natural language, hence allowing for specifying expressions more conveniently. To model the mapping from text to expressions, we first construct a text-video paired talking head dataset where each video has diverse text descriptions that depict both coarse-grained emotions and fine-grained facial movements. Leveraging the proposed dataset, we introduce a CLIP-based style encoder that projects natural language-based descriptions to the representations of expressions. TalkCLIP can even infer expressions for descriptions unseen during training. TalkCLIP can also use text to modulate expression intensity and edit expressions. Extensive experiments demonstrate that TalkCLIP achieves the advanced capability of generating photo-realistic talking heads with vivid facial expressions guided by text descriptions.

Most recent popular Role-Playing Games (RPGs) allow players to create in-game characters with hundreds of adjustable parameters, including bone positions and various makeup options. Although text-driven auto-customization systems have been developed to simplify the complex process of adjusting these intricate character parameters, they are limited by their single-round generation and lack the capability for further editing and fine-tuning. In this paper, we propose an Interactive Character Editing framework (ICE) to achieve a multi-round dialogue-based refinement process. In a nutshell, our ICE offers a more user-friendly way to enable players to convey creative ideas iteratively while ensuring that created characters align with the expectations of players. Specifically, we propose an Instruction Parsing Module (IPM) that utilizes large language models (LLMs) to parse multi-round dialogues into clear editing instruction prompts in each round. To reliably and swiftly modify character control parameters at a fine-grained level, we propose a Semantic-guided Low-dimension Parameter Solver (SLPS) that edits character control parameters according to prompts in a zero-shot manner. Our SLPS first localizes the character control parameters related to the fine-grained modification, and then optimizes the corresponding parameters in a low-dimension space to avoid unrealistic results. Extensive experimental results demonstrate the effectiveness of our proposed ICE for in-game character creation and the superior editing performance of ICE. Code:https://github.com/NeteaseFuxi/ICE-Interactive-3D-Game-Character.

Deep learning approaches have provided state-of-the-art performance in many applications by relying on large and overparameterized neural networks. However, such networks are very brittle and are difficult to deploy on resource-limited platforms. Model pruning, i.e., reducing the size of the network, is a widely adopted strategy that can lead to a more robust and compact model. Many heuristics exist for model pruning, but our understanding of the pruning process remains limited due to the black-box nature of a neural network model. Empirical studies show that some heuristics improve performance whereas others can make models more brittle. This work aims to shed light on how different pruning methods alter the network's internal feature representation and the corresponding impact on model performance. To facilitate a comprehensive comparison and characterization of the high-dimensional model feature space, we introduce a visual geometric analysis of feature representations. We evaluated a set of critical geometric concepts decomposed from the commonly adopted classification loss and used them to design a visualization system to compare and highlight the impact of pruning on model performance and feature representation. The proposed tool provides an environment for an in-depth comparison of pruning methods and a comprehensive understanding of how the model responds to common data corruption. By leveraging the proposed visualization, machine learning researchers can reveal the similarities between pruning methods and redundancy in robustness evaluation benchmarks, obtain geometric insights about the differences between pruned models that achieve superior robustness performance, and identify samples that are robust or fragile to model pruning and common data corruption.

Existing audio-visual segmentation datasets typically focus on short-trimmed videos with only one pixel-map annotation for a per-second video clip. In contrast, for untrimmed videos, the sound duration, start- and end-sounding time positions, and visual deformation of audible objects vary significantly. Therefore, we observed that current AVS models trained on trimmed videos might struggle to segment sounding objects in long videos. To investigate the feasibility of grounding audible objects in videos along both temporal and spatial dimensions, we introduce the Long-Untrimmed Audio-Visual Segmentation dataset (LU-AVS), which includes precise frame-level annotations of sounding emission times and provides exhaustive mask annotations for all frames. Considering that pixel-level annotations are difficult to achieve in some complex scenes, we also provide the bounding boxes to indicate the sounding regions. Specifically, LU-AVS contains 10M mask annotations across 6.6K videos, and 11M bounding box annotations across 7K videos. Compared with the existing datasets, LU-AVS videos are on average 4~8 times longer, with the silent duration being 3~15 times greater. Furthermore, we try our best to adapt some baseline models that were originally designed for audio-visual-relevant tasks to examine the challenges of our newly curated LU-AVS. Through comprehensive evaluation, we demonstrate the challenges of LU-AVS compared to the ones containing trimmed videos. Therefore, LU-AVS provides an ideal yet challenging platform for evaluating audio-visual segmentation and localization on untrimmed long videos. The dataset is publicly available at: https://yenanliu.github.io/LU-AVS/.

While image diffusion models have made significant progress in text-driven 3D content creation, they often fail to accurately capture the intended meaning of text prompts, especially for view information. This limitation leads to the Janus problem, where multi-faced 3D models are generated under the guidance of such diffusion models. In this paper, we propose a robust high-quality 3D content generation pipeline by exploiting orthogonal-view image guidance. First, we introduce a novel 2D diffusion model that generates an image consisting of four orthogonal-view sub-images based on the given text prompt. Then, the 3D content is created using this diffusion model. Notably, the generated orthogonal-view image provides strong geometric structure priors and thus improves 3D consistency. As a result, it effectively resolves the Janus problem and significantly enhances the quality of 3D content creation. Additionally, we present a 3D synthesis fusion network that can further improve the details of the generated 3D contents. Both quantitative and qualitative evaluations demonstrate that our method surpasses previous text-to-3D techniques. Project page: https://efficientdreamer.github.io.

Text-guided 3D face synthesis has achieved remarkable results by leveraging text-to-image (T2I) diffusion models. However, most existing works focus solely on the direct gen-eration, ignoring the editing, restricting them from synthe-sizing customized 3D faces through iterative adjustments. In this paper, we propose a unified text-guided framework from face generation to editing. In the generation stage, we propose a geometry-texture decoupled generation to miti-gate the loss of geometric details caused by coupling. Be-sides, decoupling enables us to utilize the generated geom-etry as a condition for texture generation, yielding highly geometry-texture aligned results. We further employ a fine-tuned texture diffusion model to enhance texture quality in both RGB and YUV space. In the editing stage, we first em-ploy a pre-trained diffusion model to update facial geometry or texture based on the texts. To enable sequential editing, we introduce a UV domain consistency preservation reg-ularization, preventing unintentional changes to irrelevant facial attributes. Besides, we propose a self-guided consis-tency weight strategy to improve editing efficacy while pre-serving consistency. Through comprehensive experiments, we showcase our method's superiority in face synthesis. Project page: https://faceg2e.github.io/.

Large Multimodal Models (LMMs) have achieved strong performance across a range of vision and language tasks.However, their spatial reasoning capabilities are underinvestigated.In this paper, we construct a novel VQA dataset, Spatial-MM, to comprehensively study LMMs' spatial understanding and reasoning capabilities.Our analyses on object-relationship and multi-hop reasoning reveal several important findings.Firstly, bounding boxes and scene graphs, even synthetic ones, can significantly enhance LMMs' spatial reasoning.Secondly, LMMs struggle more with questions posed from the human perspective than the camera perspective about the image.Thirdly, chain of thought (CoT) prompting does not improve model performance on complex multi-hop questions involving spatial relations.Lastly, our perturbation analysis on GQA-spatial reveals that LMMs are much stronger at basic object detection than complex spatial reasoning.We believe our new benchmark dataset and in-depth analyses can spark further research on LMMs spatial reasoning.1

We tackle the problem of novel view synthesis (NVS) which aims to generate realistic images at novel views. Unlike existing works that require either costly per-scene optimization or relatively dense views, we propose AS-NeRF, a neural rendering based approach that can achieve generalizable NVS from only sparse views. Considering the inherent spatial continuity of images, we design a novel sparse-attention based auxiliary sampling module (ASM). Given a 3D point on the ray, the ASM adaptively attends to a sparse set of view-specific 2D auxiliary locations around the 3D point’s original projection pixels, and dynamically computes the attention weights in a cross-attention manner. This enables our model to effectively exploit the local correlation among neighboring pixels, obtaining the enhanced features with more powerful representation. Extensive experiments show that our method outperforms the state-of-the-art on both real and synthetic data.

Existing plant disease classification models have achieved remarkable performance in recognizing in-laboratory diseased images. However, their performance often significantly degrades in classifying in-the-wild images. Furthermore, we observed that in-the-wild plant images may exhibit similar appearances across various diseases (i.e., small inter-class discrepancy) while the same diseases may look quite different (i.e., large intra-class variance). Motivated by this observation, we propose an in-the-wild multimodal plant disease recognition dataset that contains the largest number of disease classes but also text-based descriptions for each disease. Particularly, the newly provided text descriptions are introduced to provide rich information in textual modality and facilitate in-the-wild disease classification with small inter-class discrepancy and large intra-class variance issues. Therefore, our proposed dataset can be regarded as an ideal testbed for evaluating disease recognition methods in the real world. In addition, we further present a strong yet versatile baseline that models text descriptions and visual data through multiple prototypes for a given class. By fusing the contributions of multimodal prototypes in classification, our baseline can effectively address the small inter-class discrepancy and large intra-class variance issues. Remarkably, our baseline model can not only classify diseases but also recognize diseases in few-shot or training-free scenarios. Extensive benchmarking results demonstrate that our proposed in-the-wild multimodal dataset sets many new challenges to the plant disease recognition task and there is a large space to improve for future works.

Pre-trained vision-language models (VLMs) such as CLIP have shown excellent performance for zero-shot classification. Based on CLIP, recent methods design various learnable prompts to evaluate the zero-shot generalization capability on a base-to-novel setting. This setting assumes test samples are already divided into either base or novel classes, limiting its application to realistic scenarios. In this paper, we focus on a more challenging and practical setting: generalized zero-shot learning (GZSL), i.e., testing with no information about the base/novel division. To address this challenging zero-shot problem, we introduce two unique designs that enable us to classify an image without the need of knowing whether it comes from seen or unseen classes. Firstly, most existing methods only adopt a single latent space to align visual and linguistic features, which has a limited ability to represent complex visual-linguistic patterns, especially for fine-grained tasks. Instead, we propose a dual-space feature alignment module that effectively augments the latent space with a novel attribute space induced by a well-devised attribute reservoir. In particular, the attribute reservoir consists of a static vocabulary and learnable tokens complementing each other for flexible control over feature granularity. Secondly, finetuning CLIP models (e.g., prompt learning) on seen base classes usually sacrifices the model's original generalization capability on unseen novel classes. To mitigate this issue, we present a new topology-preserving objective that can enforce feature topology structures of the combined base and novel classes to resemble the topology of CLIP. In this manner, our model will inherit the generalization ability of CLIP through maintaining the pairwise class angles in the attribute space. Extensive experiments on twelve object recognition datasets demonstrate that our model, termed Topology-Preserving Reservoir (TPR), outperforms strong baselines including both prompt learning and conventional generative-based zero-shot methods.

Class-agnostic object detection (OD) can be a cornerstone or a bottleneck for many downstream vision tasks. Despite considerable advancements in bottom-up and multi-object discovery methods that leverage basic visual cues to identify salient objects, consistently achieving a high recall rate remains difficult due to the diversity of object types and their contextual complexity. In this work, we investigate using vision-language models (VLMs) to enhance object detection via a self-supervised prompt learning strategy. Our initial findings indicate that manually crafted text queries often result in undetected objects, primarily because detection confidence diminishes when the query words exhibit semantic overlap. To address this, we propose a Dispersing Prompt Expansion (DiPEx) approach. DiPEx progressively learns to expand a set of distinct, non-overlapping hyperspherical prompts to enhance recall rates, thereby improving performance in downstream tasks such as out-of-distribution OD. Specifically, DiPEx initiates the process by self-training generic parent prompts and selecting the one with the highest semantic uncertainty for further expansion. The resulting child prompts are expected to inherit semantics from their parent prompts while capturing more fine-grained semantics. We apply dispersion losses to ensure high inter-class discrepancy among child prompts while preserving semantic consistency between parent-child prompt pairs. To prevent excessive growth of the prompt sets, we utilize the maximum angular coverage (MAC) of the semantic space as a criterion for early termination. We demonstrate the effectiveness of DiPEx through extensive class-agnostic OD and OOD-OD experiments on MS-COCO and LVIS, surpassing other prompting methods by up to 20.1% in AR and achieving a 21.3% AP improvement over SAM.

Isolated Sign Language Recognition (ISLR) focuses on identifying individual sign language glosses. Considering the diversity of sign languages across geographical regions, developing region-specific ISLR datasets is crucial for supporting communication and research. Auslan, as a sign language specific to Australia, still lacks a dedicated large-scale word-level dataset for the ISLR task. To fill this gap, we curate \underline{\textbf{the first}} large-scale Multi-view Multi-modal Word-Level Australian Sign Language recognition dataset, dubbed MM-WLAuslan. Compared to other publicly available datasets, MM-WLAuslan exhibits three significant advantages: (1) the largest amount of data, (2) the most extensive vocabulary, and (3) the most diverse of multi-modal camera views. Specifically, we record 282K+ sign videos covering 3,215 commonly used Auslan glosses presented by 73 signers in a studio environment.Moreover, our filming system includes two different types of cameras, i.e., three Kinect-V2 cameras and a RealSense camera. We position cameras hemispherically around the front half of the model and simultaneously record videos using all four cameras. Furthermore, we benchmark results with state-of-the-art methods for various multi-modal ISLR settings on MM-WLAuslan, including multi-view, cross-camera, and cross-view. Experiment results indicate that MM-WLAuslan is a challenging ISLR dataset, and we hope this dataset will contribute to the development of Auslan and the advancement of sign languages worldwide. All datasets and benchmarks are available at MM-WLAuslan.

Video-driven 3D facial animation transfer aims to drive avatars to reproduce the expressions of actors. Existing methods have achieved remarkable results by constraining both geometric and perceptual consistency. However, geometric constraints (like those designed on facial landmarks) are insufficient to capture subtle emotions, while expression features trained on classification tasks lack fine granularity for complex emotions. To address this, we propose \textbf{FreeAvatar}, a robust facial animation transfer method that relies solely on our learned expression representation. Specifically, FreeAvatar consists of two main components: the expression foundation model and the facial animation transfer model. In the first component, we initially construct a facial feature space through a face reconstruction task and then optimize the expression feature space by exploring the similarities among different expressions. Benefiting from training on the amounts of unlabeled facial images and re-collected expression comparison dataset, our model adapts freely and effectively to any in-the-wild input facial images. In the facial animation transfer component, we propose a novel Expression-driven Multi-avatar Animator, which first maps expressive semantics to the facial control parameters of 3D avatars and then imposes perceptual constraints between the input and output images to maintain expression consistency. To make the entire process differentiable, we employ a trained neural renderer to translate rig parameters into corresponding images. Furthermore, unlike previous methods that require separate decoders for each avatar, we propose a dynamic identity injection module that allows for the joint training of multiple avatars within a single network.

Learning from bounding-boxes annotations has shown great potential in weakly-supervised 3D point cloud instance segmentation. However, we observed that existing methods would suffer severe performance degradation with perturbed bounding box annotations. To tackle this issue, we propose a complementary image prompt-induced weakly-supervised point cloud instance segmentation (CIP-WPIS) method. CIP-WPIS leverages pretrained knowledge embedded in the 2D foundation model SAM and 3D geometric prior to achieve accurate point-wise instance labels from the bounding box annotations. Specifically, CIP-WPIS first selects image views in which 3D candidate points of an instance are fully visible. Then, we generate complementary background and foreground prompts from projections to obtain SAM 2D instance mask predictions. According to these, we assign the confidence values to points indicating the likelihood of points belonging to the instance. Furthermore, we utilize 3D geometric homogeneity provided by superpoints to decide the final instance label assignments. In this fashion, we achieve high-quality 3D point-wise instance labels. Extensive experiments on both Scannet-v2 and S3DIS benchmarks proves that our method not only achieves state-of-the-art performance for bounding-boxes supervised point cloud instance segmentation, but also exhibits robustness against noisy 3D bounding-box annotations.

With the development of various generative models, misinformation in news media becomes more deceptive and easier to create, posing a significant problem. However, existing datasets for misinformation study often have limited modalities, constrained sources, and a narrow range of topics. These limitations make it difficult to train models that can effectively combat real-world misinformation. To address this, we propose a comprehensive, large-scale Multimodal Misinformation dataset for Media Authenticity Analysis ( M 3 A ), featuring broad sources and fine-grained annotations for topics and sentiments. To curate M 3 A , we collect genuine news content from 60 renowned news outlets worldwide and generate fake samples using multiple techniques. These include altering named entities in texts, swapping modalities between samples, creating new modalities, and misrepresenting movie content as news. M 3 A contains 708K genuine news samples and over 6M fake news samples, spanning text, images, audio, and video. M 3 A provides detailed multi-class labels, crucial for various misinformation detection tasks, including out-of-context detection and deepfake detection. For each task, we offer extensive benchmarks using state-of-the-art models, aiming to enhance the development of robust misinformation detection systems. • We present M 3 A , a large-scale multimodal misinformation dataset with diverse news samples. • M 3 A includes texts, images, audio, and videos from multiple reputable news outlets. • M 3 A addresses limitations in existing datasets in misinformation generation methods and scale. • We provide multi-class annotations in M 3 A for various key tasks in misinformation detection. • We propose benchmarks for M 3 A using state-of-the-art models and out-of-distribution testing.

Current face recognition models trained on large-scale face datasets have achieved promising performance. However, using face images to train a face recognition model without consent would lead to severe privacy and ethical issues. Moreover, existing face recognition models also exhibit uneven performance on different races, thus perplexing vulnerable populations. To address the aforementioned two issues, this work investigates an ethics-aware face recognition method and examines whether we can leverage synthesized faces to achieve a high-accuracy racial balanced recognition model. In a nutshell, we introduce a race-controllable and identity-innumerable face synthesis approach to generate synthetic face images, and then employ the synthesized images to improve face recognition accuracy and mitigate recognition imbalance among different races despite the scarcity of consenting images (less than 100 individuals). More importantly, the synthetic data enable us to analyze the potential impacts of races on face recognition models quantitatively and facilitate the eradication of racial imbalance in face recognition. Extensive experiments demonstrate that employing our synthetic face data improves face recognition accuracy by a large margin while mitigating the recognition imbalance across different race groups.

Malicious image tampering refers to intentionally manipulating images to make them harmful to the owners or users. It has become one of the most severe challenges to image authenticity. Conventional methods for detecting tampering by identifying visual artifacts and distortions have limitations due to the rapid advancement of image manipulation techniques, which leave fewer detectable traces. To address these challenges, we propose a proactive media authentication method using deep learning-based semi-fragile watermarks. The designed scheme utilizes deep neural networks to embed an invisible watermark into a target image that is pixel-by-pixel entangled with it, which acts as an indicator of tampering trails. Once the watermarked image is counterfeited, the embedded watermark will exhibit changes accordingly, so we can locate the tampered regions by comparing retrieved and original watermarks. This proactive authentication mechanism makes our method effective against various image tamper techniques, including image copy&move, splicing and in-painting. Although our watermark is designed to be fragile to malicious tampering operations, it remains robust to benign image-processing operations such as JPEG compression, scaling, saturation, contrast adjustments, etc. This design enables our watermark to retain effectiveness when shared over the internet. Extensive experiments demonstrate that our method achieves state-of-the-art forgery detection with superior robustness, imperceptibility and security performance.

Abstract Marker‐based optical motion capture (MoCap) aims to localize 3D human motions from a sequence of input raw markers. It is widely used to produce physical movements for virtual characters in various games such as the role‐playing game, the fighting game, and the action‐adventure game. However, the conventional MoCap cleaning and solving process is extremely labor‐intensive, time‐consuming, and usually the most costly part of game animation production. Thus, there is a high demand for automated algorithms to replace costly manual operations and achieve accurate MoCap cleaning and solving in the game industry. In this article, we design a divide‐and‐conquer‐based MoCap solving network, dubbed MarkerNet, to estimate human skeleton motions from sequential raw markers effectively. In a nutshell, our key idea is to decompose the task of direct solving of global motion from all markers into first modeling sub‐motions of local parts from the corresponding marker subsets and then aggregating sub‐motions into a global one. In this manner, our model can effectively capture local motion patterns w.r.t. different marker subsets, thus producing more accurate results compared to the existing methods. Extensive experiments on both real and synthetic data verify the effectiveness of the proposed method.

Individuals have unique facial expression and head pose styles that reflect their personalized speaking styles. Existing one-shot talking head methods cannot capture such personalized characteristics and therefore fail to produce diverse speaking styles in the final videos. To address this challenge, we propose a one-shot style-controllable talking face generation method that can obtain speaking styles from reference speaking videos and drive the one-shot portrait to speak with the reference speaking styles and another piece of audio. Our method aims to synthesize the style-controllable coefficients of a 3D Morphable Model (3DMM), including facial expressions and head movements, in a unified framework. Specifically, the proposed framework first leverages a style encoder to extract the desired speaking styles from the reference videos and transform them into style codes. Then, the framework uses a style-aware decoder to synthesize the coefficients of 3DMM from the audio input and style codes. During decoding, our framework adopts a two-branch architecture, which generates the stylized facial expression coefficients and stylized head movement coefficients, respectively. After obtaining the coefficients of 3DMM, an image renderer renders the expression coefficients into a specific person's talking-head video. Extensive experiments demonstrate that our method generates visually authentic talking head videos with diverse speaking styles from only one portrait image and an audio clip.

Since Facial Action Unit (AU) annotations require domain expertise, common AU datasets only contain a limited number of subjects. As a result, a crucial challenge for AU detection is addressing identity overfitting. We find that AUs and facial expressions are highly associated, and existing facial expression datasets often contain a large number of identities. In this paper, we aim to utilize the expression datasets without AU labels to facilitate AU detection. Specifically, we develop a novel AU detection framework aided by the Global-Local facial Expressions Embedding, dubbed GLEE-Net. Our GLEE-Net consists of three branches to extract identity-independent expression features for AU detection. We introduce a global branch for modeling the overall facial expression while eliminating the impacts of identities. We also design a local branch focusing on specific local face regions. The combined output of global and local branches is firstly pre-trained on an expression dataset as an identity-independent expression embedding, and then finetuned on AU datasets. Therefore, we significantly alleviate the issue of limited identities. Furthermore, we introduce a 3D global branch that extracts expression coefficients through 3D face reconstruction to consolidate 2D AU descriptions. Finally, a Transformer-based multi-label classifier is employed to fuse all the representations for AU detection. Extensive experiments demonstrate that our method significantly outperforms the state-of-the-art on the widely-used DISFA, BP4D and BP4D+ datasets.

Model inversion attacks involve reconstructing the training data of a target model, which raises serious privacy concerns for machine learning models. However, these attacks, especially learning-based methods, are likely to suffer from low attack accuracy, i.e., low classification accuracy of these reconstructed data by machine learning classifiers. Recent studies showed an alternative strategy of model inversion attacks, GAN-based optimization, can improve the attack accuracy effectively. However, these series of GAN-based attacks reconstruct only class-representative training data for a class, whereas learning-based attacks can reconstruct diverse data for different training data in each class. Hence, in this paper, we propose a new training paradigm for a learning-based model inversion attack that can achieve higher attack accuracy in a black-box setting. First, we regularize the training process of the attack model with an added semantic loss function and, second, we inject adversarial examples into the training data to increase the diversity of the class-related parts (i.e., he essential features for classification tasks) in training data. This scheme guides the attack model to pay more attention to the class-related parts of the original data during the data reconstruction process. The experimental results show that our method greatly boosts the performance of existing learning-based model inversion attacks. Even when no extra queries to the target model are allowed, the approach can still improve the attack accuracy of reconstructed data. This new attack shows that the severity of the threat from learning-based model inversion adversaries is underestimated and more robust defenses are required.

In this work, we propose a new Dual Min-Max Games (DMMG) based self-supervised skeleton action recognition method by augmenting unlabeled data in a contrastive learning framework. Our DMMG consists of a viewpoint variation min-max game and an edge perturbation min-max game. These two min-max games adopt an adversarial paradigm to perform data augmentation on the skeleton sequences and graph-structured body joints, respectively. Our viewpoint variation min-max game focuses on constructing various hard contrastive pairs by generating skeleton sequences from various viewpoints. These hard contrastive pairs help our model learn representative action features, thus facilitating model transfer to downstream tasks. Moreover, our edge perturbation min-max game specializes in building diverse hard contrastive samples through perturbing connectivity strength among graph-based body joints. The connectivity-strength varying contrastive pairs enable the model to capture minimal sufficient information of different actions, such as representative gestures for an action while preventing the model from overfitting. By fully exploiting the proposed DMMG, we can generate sufficient challenging contrastive pairs and thus achieve discriminative action feature representations from unlabeled skeleton data in a self-supervised manner. Extensive experiments demonstrate that our method achieves superior results under various evaluation protocols on widely-used NTU-RGB+D, NTU120-RGB+D and PKU-MMD datasets.

Existing volumetric neural rendering techniques, such as Neural Radiance Fields (NeRF), face limitations in synthesizing high-quality novel views when the camera poses of input images are imperfect. To address this issue, we propose a novel 3D reconstruction framework that enables simultaneous optimization of camera poses, dubbed CBARF (Cascaded Bundle-Adjusting NeRF). In a nutshell, our framework optimizes camera poses in a coarse-to-fine manner and then reconstructs scenes based on the rectified poses. It is observed that the initialization of camera poses has a significant impact on the performance of bundle-adjustment (BA). Therefore, we cascade multiple BA modules at different scales to progressively improve the camera poses. Meanwhile, we develop a neighbor-replacement strategy to further optimize the results of BA in each stage. In this step, we introduce a novel criterion to effectively identify poorly estimated camera poses. Then we replace them with the poses of neighboring cameras, thus further eliminating the impact of inaccurate camera poses. Once camera poses have been optimized, we employ a density voxel grid to generate high-quality 3D reconstructed scenes and images in novel views. Experimental results demonstrate that our CBARF model achieves state-of-the-art performance in both pose optimization and novel view synthesis, especially in the existence of large camera pose noise.

Given an audio-visual pair, audio-visual segmentation (AVS) aims to locate sounding sources by predicting pixel-wise maps. Previous methods assume that each sound component in an audio signal always has a visual counterpart in the image. However, this assumption overlooks that off-screen sounds and background noise often contaminate the audio recordings in real-world scenarios. They impose significant challenges on building a consistent semantic mapping between audio and visual signals for AVS models and thus impede precise sound localization. In this work, we propose a two-stage bootstrapping audio-visual segmentation framework by incorporating multi-modal foundation knowledge$^{1}$In a nutshell, our BAVS is designed to eliminate the interference of background noise or off-screen sounds in segmentation by establishing the audio-visual correspondences in an explicit manner. In the first stage, we employ a segmentation model to localize potential sounding objects from visual data without being affected by contaminated audio signals. Meanwhile, we also utilize a foundation audio classification model to discern audio semantics. Considering the audio tags provided by the audio foundation model are noisy, associating object masks with audio tags is not trivial. Thus, in the second stage, we develop an audio-visual semantic integration strategy (AVIS) to localize the authentic-sounding objects. Here, we construct an audio-visual tree based on the hierarchical correspondence between sounds and object categories. We then examine the label concurrency between the localized objects and classified audio tags by tracing the audio-visual tree. With AVIS, we can effectively segment real-sounding objects. Extensive experiments demonstrate the superiority of our method on AVS datasets, particularly in scenarios involving background noise. Our project website ishttps://yenanliu.github.io/AVSS.github.io/.

Generating vivid and diverse 3D co-speech gestures is crucial for various applications in animating virtual avatars. While most existing methods can generate gestures from audio directly, they usually overlook thatemotionis one of the key factors of authentic co-speech gesture generation. In this work, we proposeEmotionGesture, a novel framework for synthesizing vivid and diverse emotional co-speech 3D gestures from audio. Considering emotion is often entangled with the rhythmic beat in speech audio, we first develop an Emotion-Beat Mining module (EBM) to extract the emotion and audio beat features as well as model their correlation via a transcript-based visual-rhythm alignment. Then, we propose an initial pose based Spatial-Temporal Prompter (STP) to generate future gestures from the given initial poses. STP effectively models the spatial-temporal correlations between the initial poses and the future gestures, thus producing the spatial-temporal coherent pose prompt. Once we obtain pose prompts, emotion, and audio beat features, we will generate 3D co-speech gestures through a transformer architecture. However, considering the poses of existing datasets often contain jittering effects, this would lead to generating unstable gestures. To address this issue, we propose an effective objective function, dubbed Motion-Smooth Loss. Specifically, we model motion offset to compensate for jittering ground-truth by forcing gestures to be smooth. Last, we present an emotion-conditioned VAE to sample emotion features, enabling us to generate diverse emotional results. Extensive experiments demonstrate that our framework outperforms the state-of-the-art, achieving vivid and diverse emotional co-speech 3D gestures. Our code and dataset will be released on the project page:https://xingqunqi-lab.github.io/Emotion-Gesture-Web/

Automatic font generation is a challenging and time-consuming task, particularly in languages that consist of large amounts of characters with complicated structures. Typical component-wise font generation methods decompose the source character into components and search for them from the reference glyph set as candidate components. These candidate components are then utilized to learn the local styles of the target glyph. However, these methods overlook that the same component at different locations may have different profiles. When the candidate components locate differently from their corresponding components in the target glyph, the style of a generated glyph will look inconsistent. It is observed that for arbitrary components at two specific locations, the deformation patterns are similar. Driven by this, we present a location-aware component-deformable font generation method. Specifically, we search for candidate components and their corresponding deformative component pairs from the reference glyph set. Each deformative component pair can accurately depict how to deform the candidate component to the desired profile in the target glyph. Hence, we introduce a location-dependent deformation module to perform component warping. In this way, we significantly improve the component deformation ability. Lastly, we integrate deformed components into target glyphs while enforcing their styles to be consistent with the reference ones. Extensive experiments demonstrate that our method produces target-font consistent glyphs and outperforms the state-of-the-art on both seen and unseen fonts.

In this work, we propose a stroke-based hairstyle editing network, dubbed HairstyleNet, allowing users to conveniently change the hairstyles of an image in an interactive fashion. Different from previous works, we simplify the hairstyle editing process where users can manipulate local or entire hairstyles by adjusting the parameterized hair regions. Our HairstyleNet consists of two stages: a stroke parameterization stage and a stroke-to-hair generation stage. In the stroke parameterization stage, we first introduce parametric strokes to approximate the hair wisps, where the stroke shape is controlled by a quadratic Bézier curve and a thickness parameter. Since rendering strokes with thickness to an image is not differentiable, we opt to leverage a neural renderer to construct the mapping from stroke parameters to a stroke image. Thus, the stroke parameters can be directly estimated from hair regions in a differentiable way, enabling us to flexibly edit the hairstyles of input images. In the stroke-to-hair generation stage, we design a hairstyle refinement network that first encodes coarsely composed images of hair strokes, face, and background into latent representations and then generates high-fidelity face images with desirable new hairstyles from the latent codes. Extensive experiments demonstrate that our HairstyleNet achieves state-of-the-art performance and allows flexible hairstyle manipulation.

Different people speak with diverse personalized speaking styles. Although existing one-shot talking head methods have made significant progress in lip sync, natural facial expressions, and stable head motions, they still cannot generate diverse speaking styles in the final talking head videos. To tackle this problem, we propose a one-shot style-controllable talking face generation framework. In a nutshell, we aim to attain a speaking style from an arbitrary reference speaking video and then drive the one-shot portrait to speak with the reference speaking style and another piece of audio. Specifically, we first develop a style encoder to extract dynamic facial motion patterns of a style reference video and then encode them into a style code. Afterward, we introduce a style-controllable decoder to synthesize stylized facial animations from the speech content and style code. In order to integrate the reference speaking style into generated videos, we design a style-aware adaptive transformer, which enables the encoded style code to adjust the weights of the feed-forward layers accordingly. Thanks to the style-aware adaptation mechanism, the reference speaking style can be better embedded into synthesized videos during decoding. Extensive experiments demonstrate that our method is capable of generating talking head videos with diverse speaking styles from only one portrait image and an audio clip while achieving authentic visual effects. Project Page: https://github.com/FuxiVirtualHuman/styletalk.

In this work, we propose a semantic flow-guided two-stage framework for shape-aware face swapping, namely FlowFace. Unlike most previous methods that focus on transferring the source inner facial features but neglect facial contours, our FlowFace can transfer both of them to a target face, thus leading to more realistic face swapping. Concretely, our FlowFace consists of a face reshaping network and a face swapping network. The face reshaping network addresses the shape outline differences between the source and target faces. It first estimates a semantic flow (i.e. face shape differences) between the source and the target face, and then explicitly warps the target face shape with the estimated semantic flow. After reshaping, the face swapping network generates inner facial features that exhibit the identity of the source face. We employ a pre-trained face masked autoencoder (MAE) to extract facial features from both the source face and the target face. In contrast to previous methods that use identity embedding to preserve identity information, the features extracted by our encoder can better capture facial appearances and identity information. Then, we develop a cross-attention fusion module to adaptively fuse inner facial features from the source face with the target facial attributes, thus leading to better identity preservation. Extensive quantitative and qualitative experiments on in-the-wild faces demonstrate that our FlowFace outperforms the state-of-the-art significantly.

Object-goal visual navigation aims at steering an agent toward an object via a series of moving steps. Previous works mainly focus on learning informative visual representations for navigation, but overlook the impacts of navigation states on the effectiveness and efficiency of navigation. We observe that high relevance among navigation states will cause navigation inefficiency or failure for existing methods. In this paper, we present a History-inspired Navigation Policy Learning (HiNL) framework to estimate navigation states effectively by exploring relationships among historical navigation states. In HiNL, we propose a History-aware State Estimation (HaSE) module to alleviate the impacts of dominant historical states on the current state estimation. Meanwhile, HaSE also encourages an agent to be alert to the current observation changes, thus enabling the agent to make valid actions. Furthermore, we design a History-based State Regularization (HbSR) to explicitly suppress the correlation among navigation states in training. As a result, our agent can update states more effectively while reducing the correlations among navigation states. Experiments on the artificial platform AI2-THOR (i.e., iTHOR and RoboTHOR) demonstrate that HiNL significantly outperforms state-of-the-art methods on both Success Rate and SPL in unseen testing environments.

Predicting natural and diverse 3D hand gestures from the upper body dynamics is a practical yet challenging task in virtual avatar creation. Previous works usually overlook the asymmetric motions between two hands and generate two hands in a holistic manner, leading to unnatural results. In this work, we introduce a novel bilateral hand disentanglement based two-stage 3D hand generation method to achieve natural and diverse 3D hand prediction from body dynamics. In the first stage, we intend to generate natural hand gestures by two hand-disentanglement branches. Considering the asymmetric gestures and motions of two hands, we introduce a Spatial-Residual Memory (SRM) module to model spatial interaction between the body and each hand by residual learning. To enhance the coordination of two hand motions wrt. body dynamics holistically, we then present a Temporal-Motion Memory (TMM) module. TMM can effectively model the temporal association between body dynamics and two hand motions. The second stage is built upon the insight that 3D hand predictions should be non-deterministic given the sequential body postures. Thus, we further diversify our 3D hand predictions based on the initial output from the stage one. Concretely, we propose a Prototypical-Memory Sampling Strategy (PSS) to generate the non-deterministic hand gestures by gradient-based Markov Chain Monte Carlo (MCMC) sampling. Extensive experiments demonstrate that our method outperforms the state-of-the-art models on the B2H dataset and our newly collected TED Hands dataset. The dataset and code are available at: https://github.com/XingqunQilab/Diverse-3D-Hand-Gesture-Prediction.

Inverse rendering methods aim to estimate geometry, materials and illumination from multi-view RGB images. In order to achieve better decomposition, recent approaches attempt to model indirect illuminations reflected from different materials via Spherical Gaussians (SG), which, however, tends to blur the high-frequency reflection details. In this paper, we propose an end-to-end inverse rendering pipeline that decomposes materials and illumination from multi-view images, while considering near-field indirect illumination. In a nutshell, we introduce the Monte Carlo sampling based path tracing and cache the indirect illumination as neural radiance, enabling a physics-faithful and easy-to-optimize inverse rendering method. To enhance efficiency and practicality, we leverage SG to represent the smooth environment illuminations and apply importance sampling techniques. To supervise indirect illuminations from unobserved directions, we develop a novel radiance consistency constraint between implicit neural radiance and path tracing results of unobserved rays along with the joint optimization of materials and illuminations, thus significantly improving the decomposition performance. Extensive experiments demonstrate that our method outperforms the state-of-the-art on multiple synthetic and real datasets, especially in terms of inter-reflection decomposition.

Most gait recognition methods exploit spatial-temporal representations from static appearances and dynamic walking patterns. However, we observe that many part-based methods neglect representations at boundaries. In addition, the phenomenon of overfitting on training data is relatively common in gait recognition, which is perhaps due to insufficient data and low-informative gait silhouettes. Motivated by these observations, we propose a novel mask-based regularization method named ReverseMask. By injecting perturbation on the feature map, the proposed regularization method helps convolutional architecture learn the discriminative representations and enhances generalization. Also, we design an Inception-like ReverseMask Block, which has three branches composed of a global branch, a feature-dropping branch, and a feature scaling branch. Precisely, the dropping branch can extract fine-grained representations when partial activations are zero-outed. Meanwhile, the scaling branch randomly scales the feature map, keeping structural information of activations and preventing overfitting. The plug-and-play Inception-like ReverseMask block is simple and effective, improving the performance of many state-of-the-art methods. Extensive experiments demonstrate that the ReverseMask regularization help baseline achieves higher accuracy and better generalization. Moreover, the base-line with Inception-like Block significantly outperforms state-of-the-art methods on the two most popular datasets, CASIA-B and OUMVLP.

Gait recognition is a biometric technology that recognizes the identity of humans through their walking patterns. Compared with other biometric technologies, gait recognition is more difficult to disguise and can be applied to the condition of long-distance without the cooperation of subjects. Thus, it has unique potential and wide application for crime prevention and social security. At present, most gait recognition methods directly extract features from the video frames to establish representations. However, these architectures learn representations from different features equally but do not pay enough attention to dynamic features, which refers to a representation of dynamic parts of silhouettes over time (e.g. legs). Since dynamic parts of the human body are more informative than other parts (e.g. bags) during walking, in this paper, we propose a novel and high-performance framework named DyGait. This is the first framework on gait recognition that is designed to focus on the extraction of dynamic features. Specifically, to take full advantage of the dynamic information, we propose a Dynamic Augmentation Module (DAM), which can automatically establish spatial-temporal feature representations of the dynamic parts of the human body. The experimental results show that our DyGait network outperforms other state-of-the-art gait recognition methods. It achieves an average Rank-1 accuracy of 71.4% on the GREW dataset, 66.3% on the Gait3D dataset, 98.4% on the CAS1A-B dataset and 98.3% on the OU-MVLP dataset.

Existing Building Damage Detection (BDD) methods always require labour-intensive pixel-level annotations of buildings and their conditions, hence largely limiting their applications. In this paper, we investigate a challenging yet practical scenario of BDD, Unsupervised Building Damage Detection (U-BDD), where only unlabelled pre- and post-disaster satellite image pairs are provided. As a pilot study, we have first proposed an advanced U-BDD baseline that leverages pre-trained vision-language foundation models to address the U-BDD task. However, the apparent domain gap between satellite and generic images causes low confidence in the foundation models used to identify buildings and their damages. In response, we further present a novel self-supervised framework, U-BDD++, which improves upon the U-BDD baseline by addressing domain-specific issues associated with satellite imagery. Extensive experiments on the widely used building damage assessment benchmark demonstrate the effectiveness of the proposed method for unsupervised building damage detection. The presented annotation-free and foundation model-based paradigm ensures an efficient learning phase. This study opens a new direction for real-world BDD and sets a strong baseline for future research.

The audio-visual segmentation (AVS) task aims to segment sounding objects from a given video. Existing works mainly focus on fusing audio and visual features of a given video to achieve sounding object masks. However, we observed that prior arts are prone to segment a certain salient object in a video regardless of the audio information. This is because sounding objects are often the most salient ones in the AVS dataset. Thus, current AVS methods might fail to localize genuine sounding objects due to the dataset bias. In this work, we present an audio-visual instance-aware segmentation approach to overcome the dataset bias. In a nutshell, our method first localizes potential sounding objects in a video by an object segmentation network, and then associates the sounding object candidates with the given audio. We notice that an object could be a sounding object in one video but a silent one in another video. This would bring ambiguity in training our object segmentation network as only sounding objects have corresponding segmentation masks. We thus propose a silent object-aware segmentation objective to alleviate the ambiguity. Moreover, since the category information of audio is unknown, especially for multiple sounding sources, we propose to explore the audio-visual semantic correlation and then associate audio with potential objects. Specifically, we attend predicted audio category scores to potential instance masks and these scores will highlight corresponding sounding instances while suppressing inaudible ones. When we enforce the attended instance masks to resemble the ground-truth mask, we are able to establish audio-visual semantics correlation. Experimental results on the AVS benchmarks demonstrate that our method can effectively segment sounding objects without being biased to salient objects and also achieves state-of-the-art performance in both the single-source and multi-source scenarios.

Retinal vessel segmentation is generally grounded in image-based datasets collected with bench-top devices. The static images naturally lose the dynamic characteristics of retina fluctuation, resulting in diminished dataset richness, and the usage of bench-top devices further restricts dataset scalability due to its limited accessibility. Considering these limitations, we introduce the first video-based retinal dataset by employing handheld devices for data acquisition. The dataset comprises 635 smartphone-based fundus videos collected from four different clinics, involving 415 patients from 50 to 75 years old. It delivers comprehensive and precise annotations of retinal structures in both spatial and temporal dimensions, aiming to advance the landscape of vasculature segmentation. Specifically, the dataset provides three levels of spatial annotations: binary vessel masks for overall retinal structure delineation, general vein-artery masks for distinguishing the vein and artery, and fine-grained vein-artery masks for further characterizing the granularities of each artery and vein. In addition, the dataset offers temporal annotations that capture the vessel pulsation characteristics, assisting in detecting ocular diseases that require fine-grained recognition of hemodynamic fluctuation. In application, our dataset exhibits a significant domain shift with respect to data captured by bench-top devices, thus posing great challenges to existing methods. Thanks to rich annotations and data scales, our dataset potentially paves the path for more advanced retinal analysis and accurate disease diagnosis. In the experiments, we provide evaluation metrics and benchmark results on our dataset, reflecting both the potential and challenges it offers for vessel segmentation tasks. We hope this challenging dataset would significantly contribute to the development of eye disease diagnosis and early prevention.

Sign language translation (SLT) aims to convert a continuous sign language video clip into a spoken language. Considering different geographic regions generally have their own native sign languages, it is valuable to establish corresponding SLT datasets to support related communication and research. Auslan, as a sign language specific to Australia, still lacks a dedicated large-scale dataset for SLT.To fill this gap, we curate an Australian Sign Language translation dataset, dubbed Auslan-Daily, which is collected from the Auslan educational TV series and Auslan TV programs. The former involves daily communications among multiple signers in the wild, while the latter comprises sign language videos for up-to-date news, weather forecasts, and documentaries. In particular, Auslan-Daily has two main features: (1) the topics are diverse and signed by multiple signers, and (2) the scenes in our dataset are more complex, e.g., captured in various environments, gesture interference during multi-signers' interactions and various camera positions. With a collection of more than 45 hours of high-quality Auslan video materials, we invite Auslan experts to align different fine-grained visual and language pairs, including video $\leftrightarrow$ fingerspelling, video $\leftrightarrow$ gloss, and video $\leftrightarrow$ sentence. As a result, Auslan-Daily contains multi-grained annotations that can be utilized to accomplish various fundamental sign language tasks, such as signer detection, sign spotting, fingerspelling detection, isolated sign language recognition, sign language translation and alignment. Moreover, we benchmark results with state-of-the-art models for each task in Auslan-Daily. Experiments indicate that Auslan-Daily is a highly challenging SLT dataset, and we hope this dataset will contribute to the development of Auslan and the advancement of sign languages worldwide in a broader context. All datasets and benchmarks are available at Auslan-Daily.

This paper investigates how to leverage more readily acquired annotations, i.e., 3D bounding boxes instead of dense point-wise labels, for instance segmentation. We propose a Weakly-supervised point cloud Instance Segmentation framework with Geometric Priors (WISGP) that allows segmentation models to be trained with 3D bounding boxes of instances. Considering intersections among bounding boxes in a scene would result in ambiguous la- bels, we first group points into two sets, i.e., univocal and equivocal sets, indicating the certainty of a 3D point belonging to an instance, respectively. Specifically, 3D points with clear labels belong to the univocal set while the rest are grouped into the equivocal set. To assign reliable labels to points in the equivocal set, we design a Geometry-guided Label Propagation (GLP) scheme that progressively propagates labels to linked points based on geometric structure, e.g., polygon meshes and superpoints. Afterwards, we train an instance segmentation model with the univocal points and equivocal points labeled by GLP, and then employ it to assign pseudo labels for the remainder of the unlabeled points. Lastly, we retrain the model with all the labeled points to achieve better instance segmentation performance. Experiments on large-scale datasets ScanNet-v2 and S3DIS demonstrate that WISGP is superior to competing weakly-supervised algorithms and even on par with a few fully-supervised ones.

In this paper, we propose the Temporal Identity Inconsistency Network (TI2Net), a Deepfake detector that focuses on temporal identity inconsistency. Specifically, TI2Net recognizes fake videos by capturing the dissimilarities of human faces among video frames of the same identity. Therefore, TI2Net is a reference-agnostic detector and can be used on unseen datasets. For a video clip of a given identity, identity information in all frames will first be encoded to identity vectors. TI2Net learns the temporal identity embedding from the temporal difference of the identity vectors. The temporal embedding, representing the identity inconsistency in the video clip, is finally used to determine the authenticity of the video clip. During training, TI2Net incorporates triplet loss to learn more discriminative temporal embeddings. We conduct comprehensive experiments to evaluate the performance of the proposed TI2Net. Experimental results indicate that TI2Net generalizes well to unseen manipulations and datasets with unseen identities. Besides, TI2Net also shows robust performance against compression and additive noise.

Video stabilization is highly desirable when videos undergo severe jittering artifacts. The difficulty of obtaining sufficient training data obstructs the development of video stabilization. In this work, we address this issue by presenting a Sim2RealVS benchmark with more than 1,300 pairs of shaky and stable videos. Our benchmark is curated by an in-game simulator with diverse scenes and various jittering effects. Moreover, we propose a simple yet strong baseline approach, named Motion-Trajectory Smoothing Network (MTSNet), by fully exploiting our Sim2RealVS data. Our MTSNet consists of three main steps: motion estimation, global trajectory smoothing and frame warping. In motion estimation, we design a Motion Correction and Completion (MCC) module to rectify the optical flow with low confidence, such as in textureless regions, thus providing more consistent motion estimation for next steps. Benefiting from our synthetic data, we can explicitly learn a Trajectory Smoothing Transformer (TST) with ground-truth supervision to smooth global trajectories. In training TST, we propose two fully-supervised losses, i.e., a motion magnitude similarity loss and a motion tendency similarity loss. After training, our TST is able to produce smooth motion trajectories for the shaky input videos. Extensive qualitative and quantitative results demonstrate that our MTSNet achieves superior performance on both synthetic and real-world data.

The explosive progress of Deepfake techniques poses unprecedented privacy and security risks to our society by creating real-looking but fake visual content. The current Deepfake detection studies are still in their infancy because they mainly rely on capturing artifacts left by a Deepfake synthesis process as detection clues, which can be easily removed by various distortions (e.g. blurring) or advanced Deepfake techniques. In this paper, we propose a novel method that does not depend on identifying the artifacts but resorts to the mechanism of anti-counterfeit labels to protect face images from malicious Deepfake tampering. Specifically, we design a neural network with an encoder-decoder structure to embed watermarks as anti-Deepfake labels into the facial identity features. The injected label is entangled with the facial identity feature, so it will be sensitive to face swap translations (i.e., Deepfake) and robust to conventional image modifications (e.g., resize and compress). Therefore, we can identify whether watermarked images have been tampered with by Deepfake methods according to the label’s existence. Experimental results demonstrate that our method can achieve average detection accuracy of more than 80%, which validates the proposed method’s effectiveness in implementing Deepfake detection.

We address the problem of ground-to-satellite image geo-localization, that is, estimating the camera latitude, longitude and orientation (azimuth angle) by matching a query image captured at the ground level against a large-scale database with geotagged satellite images. Our prior arts treat the above task as pure image retrieval by selecting the most similar satellite reference image matching the ground-level query image. However, such an approach often produces coarse location estimates because the geotag of the retrieved satellite image only corresponds to the image center while the ground camera can be located at any point within the image. To further consolidate our prior research finding, we present a novel geometry-aware geo-localization method. Our new method is able to achieve the fine-grained location of a query image, up to pixel size precision of the satellite image, once its coarse location and orientation have been determined. Moreover, we propose a new geometry-aware image retrieval pipeline to improve the coarse localization accuracy. Apart from a polar transform in our conference work, this new pipeline also maps satellite image pixels to the ground-level plane in the ground-view via a geometry-constrained projective transform to emphasize informative regions, such as road structures, for cross-view geo-localization. Extensive quantitative and qualitative experiments demonstrate the effectiveness of our newly proposed framework. We also significantly improve the performance of coarse localization results compared to the state-of-the-art in terms of location recalls.

Single image blind deblurring is highly ill-posed as neither the latent sharp image nor the blur kernel is known. Even though considerable progress has been made, several major difficulties remain for blind deblurring, including the trade-off between high-performance deblurring and real-time processing. Besides, we observe that current single image blind deblurring networks cannot further improve or stabilize the performance but significantly degrades the performance when re-deblurring is repeatedly applied. This implies the limitation of these networks in modeling an ideal deblurring process. In this work, we make two contributions to tackle the above difficulties: (1) We introduce the idempotent constraint into the deblurring framework and present a deep idempotent network to achieve improved blind non-uniform deblurring performance with stable re-deblurring. (2) We propose a simple yet efficient deblurring network with lightweight encoder-decoder units and a recurrent structure that can deblur images in a progressive residual fashion. Extensive experiments on synthetic and realistic datasets prove the superiority of our proposed framework. Remarkably, our proposed network is nearly$6.5\times $smaller and$6.4\times $faster than the state-of-the-art while achieving comparable high performance.

Weakly-supervised object detection (WSOD), which requires only image-level annotations for training detectors, has gained enormous attention. Despite recent rapid advance in WSOD, there remains a large performance gap compared with fully-supervised object detection. To narrow the performance gap, we study cross-supervised object detection (CSOD), where existing classes (base classes) have instance-level annotations while newly added classes (novel classes) only need image-level annotations. For improving localization accuracy, we propose a Cyclic Self-Training (CST) method to introduce instance-level supervision into a commonly used WSOD method, online instance classifier refinement (OICR). Our proposed CST consists of forward pseudo labeling and backward pseudo labeling. Specifically, OICR exploits the forward pseudo labeling to generate pseudo ground-truth bounding-boxes for all classes, thus enabling instance classifier training. Then, the backward pseudo labeling is designed to generate pseudo ground-truth bounding-boxes of higher quality for novel classes by fusing the predictions of the instance classifiers. As a result, both novel and base classes will have bounding-box annotations for training, alleviating the supervision inconsistency between base and novel classes. In the forward pseudo labeling, the generated pseudo ground-truths may be misaligned with objects and thus introduce poor-quality examples for training the ICs. To reduce the impacts of these poor-quality training examples, we propose a Proposal Weight Modulation (PWM) module learned in a class-agnostic and contrastive manner by exploiting bounding-box annotations of base classes. Experiments on PASCAL VOC and MS COCO datasets demonstrate the superiority of our proposed method.

For a monocular camera-based navigation system, if we could effectively explore scene geometric cues from RGB images, the geometry information will significantly facilitate the efficiency of the navigation system. Motivated by this, we propose a highly efficient point-goal navigation framework, dubbed Geo-Nav. In a nutshell, our Geo-Nav consists of two parts: a visual perception part and a navigation part. In the visual perception part, we firstly propose a Self-supervised Depth Estimation network (SDE) specially tailored for the monocular camera-based navigation agent. Our SDE learns a mapping from an RGB input image to its corresponding depth image by exploring scene geometric constraints in a self-consistency manner. Then, in order to achieve a representative visual representation from the RGB inputs and learned depth images, we propose a Cross-modality Pyramid Fusion module (CPF). Concretely, our CPF computes a patch-wise cross-modality correlation between different modal features and exploits the correlation to fuse and enhance features at each scale. Thanks to the patch-wise nature of our CPF, we can fuse feature maps at high resolution, allowing our visual network to perceive more image details. In the navigation part, our extracted visual representations are fed to a navigation policy network to learn how to map the visual representations to agent actions effectively. Extensive experiments on a widely-used multiple-room environment Gibson demonstrate that Geo-Nav outperforms the state-of-the-art in terms of efficiency and effectiveness.

Audio-driven one-shot talking face generation methods are usually trained on video resources of various persons. However, their created videos often suffer unnatural mouth shapes and asynchronous lips because those methods struggle to learn a consistent speech style from different speakers. We observe that it would be much easier to learn a consistent speech style from a specific speaker, which leads to authentic mouth movements. Hence, we propose a novel one-shot talking face generation framework by exploring consistent correlations between audio and visual motions from a specific speaker and then transferring audio-driven motion fields to a reference image. Specifically, we develop an Audio-Visual Correlation Transformer (AVCT) that aims to infer talking motions represented by keypoint based dense motion fields from an input audio. In particular, considering audio may come from different identities in deployment, we incorporate phonemes to represent audio signals. In this manner, our AVCT can inherently generalize to audio spoken by other identities. Moreover, as face keypoints are used to represent speakers, AVCT is agnostic against appearances of the training speaker, and thus allows us to manipulate face images of different identities readily. Considering different face shapes lead to different motions, a motion field transfer module is exploited to reduce the audio-driven dense motion field gap between the training identity and the one-shot reference. Once we obtained the dense motion field of the reference image, we employ an image renderer to generate its talking face videos from an audio clip. Thanks to our learned consistent speaking style, our method generates authentic mouth shapes and vivid movements. Extensive experiments demonstrate that our synthesized videos outperform the state-of-the-art in terms of visual quality and lip-sync.

In this paper, we present a novel double diffusion based neural radiance field, dubbed DD-NeRF, to reconstruct human body geometry and render the human body appearance in novel views from a sparse set of images. We first propose a double diffusion mechanism to achieve expressive representations of input images by fully exploiting human body priors and image appearance details at two levels. At the coarse level, we first model the coarse human body poses and shapes via an unclothed 3D deformable vertex model as guidance. At the fine level, we present a multi-view sampling network to capture subtle geometric deformations and image detailed appearances, such as clothing and hair, from multiple input views. Considering the sparsity of the two level features, we diffuse them into feature volumes in the canonical space to construct neural radiance fields. Then, we present a signed distance function (SDF) regression network to construct body surfaces from the diffused features. Thanks to our double diffused representations, our method can even synthesize novel views of unseen subjects. Experiments on various datasets demonstrate that our approach outperforms the state-of-the-art in both geometric reconstruction and novel view synthesis.

In point cloud videos, point coordinates are irregular and unordered but point timestamps exhibit regularities and order. Grid-based networks for conventional video processing cannot be directly used to model raw point cloud videos. Therefore, in this work, we propose a point-based network that directly handles raw point cloud videos. First, to preserve the spatio-temporal local structure of point cloud videos, we design a point tube covering a local range along spatial and temporal dimensions. By progressively subsampling frames and points and enlarging the spatial radius as the point features are fed into higher-level layers, the point tube can capture video structure in a spatio-temporally hierarchical manner. Second, to reduce the impact of the spatial irregularity on temporal modeling, we decompose space and time when extracting point tube representations. Specifically, a spatial operation is employed to encode the local structure of each spatial region in a tube and a temporal operation is used to encode the dynamics of the spatial regions along the tube. Empirically, the proposed network shows strong performance on 3D action recognition, 4D semantic segmentation and scene flow estimation. Theoretically, we analyse the necessity to decompose space and time in point cloud video modeling and why the network outperforms existing methods.

Event-based cameras measure intensity changes (called 'events') with microsecond accuracy under high-speed motion and challenging lighting conditions. With the 'active pixel sensor' (APS), the 'Dynamic and Active-pixel Vision Sensor' (DAVIS) allows the simultaneous output of intensity frames and events. However, the output images are captured at a relatively low frame rate and often suffer from motion blur. A blurred image can be regarded as the integral of a sequence of latent images, while events indicate changes between the latent images. Thus, we are able to model the blur-generation process by associating event data to a latent sharp image. Based on the abundant event data alongside a low frame rate, easily blurred images, we propose a simple yet effective approach to reconstruct high-quality and high frame rate sharp videos. Starting with a single blurred frame and its event data from DAVIS, we propose the Event-based Double Integral (EDI) model and solve it by adding regularization terms. Then, we extend it to multiple Event-based Double Integral (mEDI) model to get more smooth results based on multiple images and their events. Furthermore, we provide a new and more efficient solver to minimize the proposed energy model. By optimizing the energy function, we achieve significant improvements in removing blur and the reconstruction of a high temporal resolution video. The video generation is based on solving a simple non-convex optimization problem in a single scalar variable. Experimental results on both synthetic and real datasets demonstrate the superiority of our mEDI model and optimization method compared to the state-of-the-art.

This paper presents a new approach for synthesizing a novel street-view panorama given a satellite image, as if captured from the geographical location at the center of the satellite image. Existing works approach this as an image generation problem, adopting generative adversarial networks to implicitly learn the cross-view transformations, but ignore the geometric constraints. In this paper, we make the geometric correspondences between the satellite and street-view images explicit so as to facilitate the transfer of information between domains. Specifically, we observe that when a 3D point is visible in both views, and the height of the point relative to the camera is known, there is a deterministic mapping between the projected points in the images. Motivated by this, we develop a novel satellite to street-view projection (S2SP) module which learns the height map and projects the satellite image to the ground-level viewpoint, explicitly connecting corresponding pixels. With these projected satellite images as input, we next employ a generator to synthesize realistic street-view panoramas that are geometrically consistent with the satellite images. Our S2SP module is differentiable and the whole framework is trained in an end-to-end manner. Extensive experimental results on two cross-view benchmark datasets demonstrate that our method generates more accurate and consistent images than existing approaches.

Existing face hallucination methods based on convolutional neural networks (CNNs) have achieved impressive performance on low-resolution (LR) faces in a normal illumination condition. However, their performance degrades dramatically when LR faces are captured in non-uniform illumination conditions. This paper proposes a Recursive Copy and Paste Generative Adversarial Network (Re-CPGAN) to recover authentic high-resolution (HR) face images while compensating for non-uniform illumination. To this end, we develop two key components in our Re-CPGAN: internal and recursive external Copy and Paste networks (CPnets). Our internal CPnet exploits facial self-similarity information residing in the input image to enhance facial details; while our recursive external CPnet leverages an external guided face for illumination compensation. Specifically, our recursive external CPnet stacks multiple external Copy and Paste (EX-CP) units in a compact model to learn normal illumination and enhance facial details recursively. By doing so, our method offsets illumination and upsamples facial details progressively in a coarse-to-fine fashion, thus alleviating the ambiguity of correspondences between LR inputs and external guided inputs. Furthermore, a new illumination compensation loss is developed to capture illumination from the external guided face image effectively. Extensive experiments demonstrate that our method achieves authentic HR face images in a uniform illumination condition with a 16× magnification factor and outperforms state-of-the-art methods qualitatively and quantitatively.

Hand-object interaction plays a very important role when humans manipulate objects. While existing methods focus on improving hand-object recognition with fully automatic methods, human intention has been largely neglected in the recognition process, thus leading to undesirable interaction descriptions. To better interpret human-object interaction that is aligned to human intention, we argue that a reference specifying human intention should be taken into account. Thus, we propose a new approach to represent interactions while reflecting human purpose with three key factors,i.e., hand, object and reference. Specifically, we design a pattern ofhand-object, object-reference, hand, object, reference> (HOR) to recognize intention based atomic hand-object interactions. This pattern aims to model interactions with the states of hand, object, reference and their relationships. Furthermore, we design a simple yet effective Spatially Part-based (3+1)D convolutional neural network, namely SP(3+1)D, which leverages 3D and 1D convolutions to model visual dynamics and object position changes based on our HOR, respectively. With the help of our SP(3+1)D network, the recognition results are able to indicate human purposes accurately. To evaluate the proposed method, we annotate a Something-1.3k dataset, which contains 10 atomic hand-object interactions and about 130 videos for each interaction. Experimental results on Something-1.3k demonstrate the effectiveness of our SP(3+1)D network.

RGB-D salient object detection (SOD) has attracted increasingly more attention as it shows more robust results in complex scenes compared with RGB SOD. However, state-of-the-art RGB-D SOD approaches heavily rely on a large amount of pixel-wise annotated data for training. Such densely labeled annotations are often labor-intensive and costly. To reduce the annotation burden, we investigate RGB-D SOD from a weakly supervised perspective. More specifically, we use annotator-friendly scribble annotations as supervision signals for model training. Since scribble annotations are much sparser compared to ground-truth masks, some critical object structure information might be neglected. To preserve such structure information, we explicitly exploit the complementary edge information from two modalities (i.e., RGB and depth). Specifically, we leverage the dual-modal edge guidance and introduce a new network architecture with a dual-edge detection module and a modality-aware feature fusion module. In order to use the useful information of unlabeled pixels, we introduce a prediction consistency training scheme by comparing the predictions of two networks optimized by different strategies. Moreover, we develop an active scribble boosting strategy to provide extra supervision signals with negligible annotation cost, leading to significant SOD performance improvement. Extensive experiments on seven benchmarks validate the superiority of our proposed method. Remarkably, the proposed method with scribble annotations achieves competitive performance in comparison to fully supervised state-of-the-art methods.

In this paper, we study the task of hallucinating an authentic high-resolution (HR) face from an occluded thumbnail. We propose a multi-stage Progressive Upsampling and Inpainting Generative Adversarial Network, dubbed Pro-UIGAN, which exploits facial geometry priors to replenish and upsample ( 8× ) the occluded and tiny faces ( 16×16 pixels). Pro-UIGAN iteratively (1) estimates facial geometry priors for low-resolution (LR) faces and (2) acquires non-occluded HR face images under the guidance of the estimated priors. Our multi-stage hallucination network upsamples and inpaints occluded LR faces via a coarse-to-fine fashion, significantly reducing undesirable artifacts and blurriness. Specifically, we design a novel cross-modal attention module for facial priors estimation, in which an input face and its landmark features are formulated as queries and keys, respectively. Such a design encourages joint feature learning across the input facial and landmark features, and deep feature correspondences will be discovered by attention. Thus, facial appearance features and facial geometry priors are learned in a mutually beneficial manner. Extensive experiments show that our Pro-UIGAN attains visually pleasing completed HR faces, thus facilitating downstream tasks, i.e., face alignment, face parsing, face recognition as well as expression classification.

Although deep neural networks have been proved effective in many applications, they are data hungry, and training deep models often requires laboriously labeled data. However, when labeled data contain erroneous labels, they often lead to model performance degradation. A common solution is to assign each sample with a dynamic weight during optimization, and the weight is adjusted in accordance with the loss. However, those weights are usually unreliable since they are measured by the losses of corrupted labels. Thus, this scheme might impede the discriminative ability of neural networks trained on noisy data. To address this issue, we propose a novel reweighting method, dubbed self-reweighting from class centroids (SRCC), by assigning sample weights based on the similarities between the samples and our online learned class centroids. Since we exploit statistical class centers in the image feature space to reweight data samples in learning, our method is robust to noise caused by corrupted labels. In addition, even after reweighting the noisy data, the decision boundaries might still suffer distortions. Thus, we leverage mixed inputs that are generated by linearly interpolating two random images and their labels to further regularize the boundaries. We employ the learned class centroids to evaluate the confidence of our generated mixed data via measuring feature similarities. During the network optimization, the class centroids are updated as more discriminative feature representations of original images are learned. In doing so, SRCC will generate more robust weighting coefficients for noisy and mixed data and facilitates our feature representation learning in return. Extensive experiments on both the synthetic and real image recognition tasks demonstrate that our method SRCC outperforms the state of the art on learning with noisy data.

Existing deraining approaches represent rain streaks with different rain layers and then separate the layers from the background image. However, because of the complexity of real-world rain, such as various densities, shapes, and directions of rain streaks, it is very difficult to decompose a rain image into clean background and rain layers. In this article, we develop a novel single-image deraining method based on residual multiscale pyramid to mitigate the difficulty of rain image decomposition. To be specific, we progressively remove rain streaks in a coarse-to-fine fashion, where heavy rain is first removed in coarse-resolution levels and then light rain is eliminated in fine-resolution levels. Furthermore, based on the observation that residuals between a restored image and its corresponding rain image give critical clues of rain streaks, we regard the residuals as an attention map to remove rains in the consecutive finer level image. To achieve a powerful yet compact deraining framework, we construct our network by recurrent layers and remove rain with the same network in different pyramid levels. In addition, we design a multiscale kernel selection network (MSKSN) to facilitate our single network to remove rain streaks at different levels. In this manner, we reduce 81% of the model parameters without decreasing deraining performance compared with our prior work. Extensive experimental results on widely used benchmarks show that our approach achieves superior deraining performance compared with the state of the art.

Existing image segmentation networks mainly leverage large-scale labeled datasets to attain high accuracy. However, labeling medical images is very expensive since it requires sophisticated expert knowledge. Thus, it is more desirable to employ only a few labeled data in pursuing high segmentation performance. In this paper, we develop a data augmentation method for one-shot brain magnetic resonance imaging (MRI) image segmentation which exploits only one labeled MRI image (named atlas) and a few unlabeled images. In particular, we propose to learn the probability distributions of deformations (including shapes and intensities) of different unlabeled MRI images with respect to the atlas via 3D variational autoencoders (VAEs). In this manner, our method is able to exploit the learned distributions of image deformations to generate new authentic brain MRI images, and the number of generated samples will be sufficient to train a deep segmentation network. Furthermore, we introduce a new standard segmentation benchmark to evaluate the generalization performance of a segmentation network through a cross-dataset setting (collected from different sources). Extensive experiments demonstrate that our method outperforms the state-of-the-art one-shot medical segmentation methods. Our code has been released at https://github.com/dyh127/Modeling-the-Probabilistic-Distribution-of-Unlabeled-Data.

In this paper, we propose a novel text-based talking-head video generation framework that synthesizes high-fidelity facial expressions and head motions in accordance with contextual sentiments as well as speech rhythm and pauses. To be specific, our framework consists of a speaker-independent stage and a speaker-specific stage. In the speaker-independent stage, we design three parallel networks to generate animation parameters of the mouth, upper face, and head from texts, separately. In the speaker-specific stage, we present a 3D face model guided attention network to synthesize videos tailored for different individuals. It takes the animation parameters as input and exploits an attention mask to manipulate facial expression changes for the input individuals. Furthermore, to better establish authentic correspondences between visual motions (i.e., facial expression changes and head movements) and audios, we leverage a high-accuracy motion capture dataset instead of relying on long videos of specific individuals. After attaining the visual and audio correspondences, we can effectively train our network in an end-to-end fashion. Extensive experiments on qualitative and quantitative results demonstrate that our algorithm achieves high-quality photo-realistic talking-head videos including various facial expressions and head motions according to speech rhythms and outperforms the state-of-the-art.

Video deblurring models exploit consecutive frames to remove blurs from camera shakes and object motions. In order to utilize neighboring sharp patches, typical methods rely mainly on homography or optical flows to spatially align neighboring blurry frames. However, such explicit approaches are less effective in the presence of fast motions with large pixel displacements. In this work, we propose a novel implicit method to learn spatial correspondence among blurry frames in the feature space. To construct distant pixel correspondences, our model builds a correlation volume pyramid among all the pixel-pairs between neigh-boring frames. To enhance the features of the reference frame, we design a correlative aggregation module that maximizes the pixel-pair correlations with its neighbors based on the volume pyramid. Finally, we feed the aggregated features into a reconstruction module to obtain the restored frame. We design a generative adversarial paradigm to optimize the model progressively. Our proposed method is evaluated on the widely-adopted DVD dataset, along with a newly collected High-Frame-Rate (1000 fps) Dataset for Video Deblurring (HFR-DVD). Quantitative and qualitative experiments show that our model performs favorably on both datasets against previous state-of-the-art methods, confirming the benefit of modeling all-range spatial correspondence for video deblurring.

Existing rain-removal algorithms often tackle either rain streak removal or raindrop removal, and thus may fail to handle real-world rainy scenes. Besides, the lack of real-world deraining datasets comprising different types of rain and their corresponding rain-free ground-truth also impedes deraining algorithm development. In this paper, we aim to address real-world deraining problems from two aspects. First, we propose a complementary cascaded network architecture, namely CCN, to remove rain streaks and raindrops in a unified framework. Specifically, our CCN removes raindrops and rain streaks in a complementary fashion, i.e., raindrop removal followed by rain streak removal and vice versa, and then fuses the results via an attention based fusion module. Considering significant shape and structure differences between rain streaks and raindrops, it is difficult to manually design a sophisticated network to remove them effectively. Thus, we employ neural architecture search to adaptively find optimal architectures within our specified deraining search space. Second, we present a new real-world rain dataset, namely RainDS, to prosper the development of deraining algorithms in practical scenarios. RainDS consists of rain images in different types and their corresponding rain-free ground-truth, including rain streak only, raindrop only, and both of them. Extensive experimental results on both existing benchmarks and RainDS demonstrate that our method outperforms the state-of-the-art.

We address the problem of novel view synthesis (NVS) from a few sparse source view images. Conventional image-based rendering methods estimate scene geometry and synthesize novel views in two separate steps. However, erroneous geometry estimation will decrease NVS performance as view synthesis highly depends on the quality of estimated scene geometry. In this paper, we propose an end-to-end NVS framework to eliminate the error propagation issue. To be specific, we construct a volume under the target view and design a source-view visibility estimation (SVE) module to determine the visibility of the target-view voxels in each source view. Next, we aggregate the visibility of all source views to achieve a consensus volume. Each voxel in the consensus volume indicates a surface existence probability. Then, we present a soft ray-casting (SRC) mechanism to find the most front surface in the target view (i.e., depth). Specifically, our SRC traverses the consensus volume along viewing rays and then estimates a depth probability distribution. We then warp and aggregate source view pixels to synthesize a novel view based on the estimated source-view visibility and target-view depth. At last, our network is trained in an end-to-end self-supervised fashion, thus significantly alleviating error accumulation in view synthesis. Experimental results demonstrate that our method generates novel views in higher quality compared to the state-of-the-art.

Compared to 2D object bounding-box labeling, it is very difficult for humans to annotate 3D object poses, especially when depth images of scenes are unavailable. This paper investigates whether we can estimate the object poses effectively when only RGB images and 2D object annotations are given. To this end, we present a two-step pose estimation framework to attain 6DoF object poses from 2D object bounding-boxes. In the first step, the framework learns to segment objects from real and synthetic data in a weakly-supervised fashion, and the segmentation masks will act as a prior for pose estimation. In the second step, we design a dual-scale pose estimation network, namely DSC-PoseNet, to predict object poses by employing a differential renderer. To be specific, our DSC-PoseNet firstly predicts object poses in the original image scale by comparing the segmentation masks and the rendered visible object masks. Then, we resize object regions to a fixed scale to estimate poses once again. In this fashion, we eliminate large scale variations and focus on rotation estimation, thus facilitating pose estimation. Moreover, we exploit the initial pose estimation to generate pseudo ground-truth to train our DSC-PoseNet in a self-supervised manner. The estimation results in these two scales are ensembled as our final pose estimation. Extensive experiments on widely-used benchmarks demonstrate that our method outperforms state-of-the-art models trained on synthetic data by a large margin and even is on par with several fully-supervised methods.

Existing RGB-D saliency detection models do not explicitly encourage RGB and depth to achieve effective multi-modal learning. In this paper, we introduce a novel multistage cascaded learning framework via mutual information minimization to explicitly model the multi-modal information between RGB image and depth data. Specifically, we first map the feature of each mode to a lower dimensional feature vector, and adopt mutual information minimization as a regularizer to reduce the redundancy between appearance features from RGB and geometric features from depth. We then perform multi-stage cascaded learning to impose the mutual information minimization constraint at every stage of the network. Extensive experiments on benchmark RGB-D saliency datasets illustrate the effectiveness of our framework. Further, to prosper the development of this field, we contribute the largest (7× larger than NJU2K) COME15K dataset, which contains 15,625 image pairs with high quality polygon-/scribble-/object-/instance-/rank-level annotations. Based on these rich labels, we additionally construct four new benchmarks with strong baselines and observe some interesting phenomena, which can motivate future model design. Source code and dataset are available at https://github.com/JingZhang617/cascaded_rgbd_sod.

Most existing brain tumor segmentation methods usually exploit multi-modal magnetic resonance imaging (MRI) images to achieve high segmentation performance. However, the problem of missing certain modality images often happens in clinical practice, thus leading to severe segmentation performance degradation. In this work, we propose a Region-aware Fusion Network (RFNet) that is able to exploit different combinations of multi-modal data adaptively and effectively for tumor segmentation. Considering different modalities are sensitive to different brain tumor regions, we design a Region-aware Fusion Module (RFM) in RFNet to conduct modal feature fusion from available image modalities according to disparate regions. Benefiting from RFM, RFNet can adaptively segment tumor regions from an incomplete set of multi-modal images by effectively aggregating modal features. Furthermore, we also develop a segmentation-based regularizer to prevent RFNet from the insufficient and unbalanced training caused by the incomplete multi-modal data. Specifically, apart from obtaining segmentation results from fused modal features, we also segment each image modality individually from the corresponding encoded features. In this manner, each modal encoder is forced to learn discriminative features, thus improving the representation ability of the fused features. Remarkably, extensive experiments on BRATS2020, BRATS2018 and BRATS2015 datasets demonstrate that our RFNet outperforms the state-of-the-art significantly.

Conventional face super-resolution methods usually assume testing low-resolution (LR) images lie in the same domain as the training ones. Due to different lighting conditions and imaging hardware, domain gaps between training and testing images inevitably occur in many real-world scenarios. Neglecting those domain gaps would lead to inferior face super-resolution (FSR) performance. However, how to transfer a trained FSR model to a target domain efficiently and effectively has not been investigated. To tackle this problem, we develop a Domain-Aware Pyramid-based Face Super-Resolution network, named DAP-FSR network. Our DAP-FSR makes the first attempt to super-resolve LR faces from a target domain by exploiting only a pair of high-resolution (HR) and LR exemplar in the target domain. To be specific, our DAP-FSR firstly employs its encoder to extract the multi-scale latent representations of the input LR face. Considering only one target domain example is available, we propose to augment the target domain data by mixing the latent representations of the target domain face and source domain ones, and then feed the mixed representations to the decoder of our DAP-FSR. The decoder will generate new face images resembling the target domain image style. The generated HR faces in turn are used to optimize our decoder to reduce the domain gap. By iteratively updating the latent representations and our decoder, our DAP-FSR will be adapted to the target domain, thus achieving authentic and high-quality upsampled HR faces. Extensive experiments on three benchmarks validate the effectiveness and superior performance of our DAP-FSR compared to the state-of-the-art methods.

Gait recognition is one of the most important biometric technologies and has been applied in many fields. Recent gait recognition frameworks represent each gait frame by descriptors extracted from either global appearances or local regions of humans. However, the representations based on global information often neglect the details of the gait frame, while local region based descriptors cannot capture the relations among neighboring regions, thus reducing their discriminativeness. In this paper, we propose a novel feature extraction and fusion framework to achieve discriminative feature representations for gait recognition. Towards this goal, we take advantage of both global visual information and local region details and develop a Global and Local Feature Extractor (GLFE). Specifically, our GLFE module is composed of our newly designed multiple global and local convolutional layers (GLConv) to ensemble global and local features in a principle manner. Furthermore, we present a novel operation, namely Local Temporal Aggregation (LTA), to further preserve the spatial information by reducing the temporal resolution to obtain higher spatial resolution. With the help of our GLFE and LTA, our method significantly improves the discriminativeness of our visual features, thus improving the gait recognition performance. Extensive experiments demonstrate that our proposed method outperforms state-of-the-art gait recognition methods on two popular datasets.

We propose PR-RRN, a novel neural-network based method for Non-rigid Structure-from-Motion (NRSfM). PR-RRN consists of Residual-Recursive Networks (RRN) and two extra regularization losses. RRN is designed to effectively recover 3D shape and camera from 2D keypoints with novel residual-recursive structure. As NRSfM is a highly under-constrained problem, we propose two new pairwise regularization to further regularize the reconstruction. The Rigidity-based Pairwise Contrastive Loss regularizes the shape representation by encouraging higher similarity between the representations of high-rigidity pairs of frames than low-rigidity pairs. We propose minimum singular-value ratio to measure the pairwise rigidity. The Pairwise Consistency Loss enforces the reconstruction to be consistent when the estimated shapes and cameras are exchanged between pairs. Our approach achieves state-of-the-art performance on CMU MOCAP and PASCAL3D+ dataset.

Multi-instance scenes are especially challenging for end-to-end visuomotor (image-to-control) learning algorithms. "Pipeline" visual servo control algorithms use separate detection, selection and servo stages, allowing algorithms to focus on a single object instance during servo control. End-to-end systems do not have separate detection and selection stages and need to address the visual ambiguities introduced by the presence of an arbitrary number of visually identical or similar objects during servo control. However, end-to-end schemes avoid embedding errors from detection and selection stages in the servo control behaviour, are more dynamically robust to changing scenes and are algorithmically simpler. In this paper, we present a reactive real-time end-to-end visuomotor learning algorithm for multi-instance reaching. The proposed algorithm uses a monocular RGB image and the manipulator’s joint angles as the input to a light-weight fully-convolutional network (FCN) to generate control candidates. A key innovation of the proposed method is identifying the optimal control candidate by regressing a control-Lyapunov function (cLf) value. The multi-instance capability emerges naturally from the stability analysis associated with the cLf formulation. We demonstrate the proposed algorithm effectively reaching and grasping objects from different categories on a table-top amid other instances and distractors from an over-the-shoulder monocular RGB camera. The network is able to run up to ∼160 fps during inference on one GTX 1080 Ti GPU.

We propose an audio-driven talking-head method to generate photo-realistic talking-head videos from a single reference image. In this work, we tackle two key challenges: (i) producing natural head motions that match speech prosody, and (ii)} maintaining the appearance of a speaker in a large head motion while stabilizing the non-face regions. We first design a head pose predictor by modeling rigid 6D head movements with a motion-aware recurrent neural network (RNN). In this way, the predicted head poses act as the low-frequency holistic movements of a talking head, thus allowing our latter network to focus on detailed facial movement generation. To depict the entire image motions arising from audio, we exploit a keypoint based dense motion field representation. Then, we develop a motion field generator to produce the dense motion fields from input audio, head poses, and a reference image. As this keypoint based representation models the motions of facial regions, head, and backgrounds integrally, our method can better constrain the spatial and temporal consistency of the generated videos. Finally, an image generation network is employed to render photo-realistic talking-head videos from the estimated keypoint based motion fields and the input reference image. Extensive experiments demonstrate that our method produces videos with plausible head motions, synchronized facial expressions, and stable backgrounds and outperforms the state-of-the-art.

Deep learning algorithms such as Convolutional Neural Networks (CNNs) are currently used to solve a range of robotics and computer vision problems. These networks typically estimate the desired representation in a single forward pass and must therefore learn to converge from a wide range of initial conditions to a precise result. This is challenging, and has led to increased interest in the development of separate refinement modules which learn to improve a given initial estimate, thus reducing the required search space. Such modules are usually developed ad-hoc for each given application, often requiring significant engineering investment. In this work we propose a generic innovation-based CNN. Our CNN is implemented along with a stochastic gradient descent (SGD) algorithm to iteratively refine a given initial estimate. The proposed approach provides a general framework for the development of refinement modules applicable to a wide range of robotics problems. We apply this framework to object pose estimation and depth estimation and demonstrate significant improvement over the initial estimates, in the range of 4.2 - 8.1%, for both applications.

The availability of a large labeled dataset is a key requirement for applying deep learning methods to solve various computer vision tasks. In the context of understanding human activities, existing public datasets, while large in size, are often limited to a single RGB camera and provide only per-frame or per-clip action annotations. To enable richer analysis and understanding of human activities, we introduce IKEA ASM-a three million frame, multi-view, furniture assembly video dataset that includes depth, atomic actions, object segmentation, and human poses. Additionally, we benchmark prominent methods for video action recognition, object segmentation and human pose estimation tasks on this challenging dataset. The dataset enables the development of holistic methods, which integrate multi-modal and multi-view data to better perform on these tasks.

Existing visual navigation approaches leverage classification neural networks to extract global features from visual data for navigation. However, these networks are not originally designed for navigation tasks. Thus, the neural architectures might not be suitable to capture scene con-tents. Fortunately, neural architecture search (NAS) brings a hope to solve this problem. In this paper, we propose an Auto-Navigator to customize a specialized network for visual navigation. However, as navigation tasks mainly rely on reinforcement learning (RL) rewards in training, such weak supervision is insufficiently indicative for NAS to optimize visual perception network. Thus, we introduce imitation learning (IL) with optimal paths to optimize navigation policies while selecting an optimal architecture. As Auto-Navigator can obtain a direct supervision in every step, such guidance greatly facilitates architecture search. In particular, we initialize our Auto-Navigator with a learnable distribution over the search space of visual perception architecture, and then optimize the distribution with IL supervision. Afterwards, we employ an RL reward function to fine-tune our Auto-Navigator to improve the generalization ability of our model. Extensive experiments demonstrate that our Auto-Navigator outperforms baseline methods on Gibson and Matterport3D without significantly increasing network parameters.

Face anti-spoofing (FAS) techniques play an important role in defending face recognition systems against spoofing attacks. Existing FAS methods often require a large number of annotated spoofing face data to train effective anti-spoofing models. Considering the attacking nature of spoofing data and its diverse variants, obtaining all the spoofing types in advance is difficult. This would limit the performance of FAS networks in practice. Thus, an online learning FAS method is highly desirable. In this paper, we present a semi-supervised learning based framework to tackle face spoofing attacks with only a few labeled training data (e.g., ∼ 50 face images). Specifically, we progressively adopt the unlabeled data with reliable pseudo labels during training to enrich the variety of training data. We observed that face spoofing data are naturally presented in the format of video streams. Thus, we exploit the temporal consistency to consolidate the reliability of a pseudo label for a selected image. Furthermore, we propose an adaptive transfer mechanism to ameliorate the influence of unseen spoofing data. Benefiting from the progressively-labeling nature of our method, we are able to train our network on not only data of seen spoofing types (i.e., the source domain) but also unlabeled data of unseen attacking types (i.e., the target domain). In this way, our method can reduce the domain gap and is more practical in real-world anti-spoofing scenarios. Extensive experiments in both the intra-database and inter-database scenarios demonstrate that our method is on par with the state-of-the-art methods but employs remarkably less labeled data (less than 0.1% labeled spoofing data in a dataset). Moreover, our method significantly outperforms fully-supervised methods on cross-domain testing scenarios with the help of our progressive learning fashion.

Weakly supervised object detection has attracted more and more attention as it only needs image-level annotations for training object detectors. A popular solution to this task is to train a multiple instance detection network (MIDN) which integrates multiple instance learning into a deep convolutional neural network. One major issue of the MIDN is that it is prone to be stuck at local discriminative regions. To address this local optimum issue, we propose a pyramidal MIDN (P-MIDN) comprised of a sequence of multiple MIDNs. In particular, one MIDN performs proposal removal for its subsequent MIDN to reduce the exposure of local discriminative proposal regions to the latter during training. In this manner, it allows our MIDNs to focus on proposals which cover objects more completely. Furthermore, we integrate the P-MIDN into an online instance classifier refinement (OICR) framework. Combined with the P-MIDN, a mask guided self-correction (MGSC) method is proposed to generate high-quality pseudo ground-truths for training the OICR. Experimental results on PASCAL VOC 2007, PASCAL VOC 2010, PASCAL VOC 2012, ILSVRC 2013 DET and MS-COCO benchmarks demonstrate that our approach achieves state-of-the-art performance.

Obtaining a high-quality frontal face image from a low-resolution (LR) non-frontal face image is primarily important for many facial analysis applications. However, mainstreams either focus on super-resolving near-frontal LR faces or frontalizing non-frontal high-resolution (HR) faces. It is desirable to perform both tasks seamlessly for daily-life unconstrained face images. In this paper, we present a novel Vivid Face Hallucination Generative Adversarial Network (VividGAN) for simultaneously super-resolving and frontalizing tiny non-frontal face images. VividGAN consists of coarse-level and fine-level Face Hallucination Networks (FHnet) and two discriminators, i.e., Coarse-D and Fine-D. The coarse-level FHnet generates a frontal coarse HR face and then the fine-level FHnet makes use of the facial component appearance prior, i.e., fine-grained facial components, to attain a frontal HR face image with authentic details. In the fine-level FHnet, we also design a facial component-aware module that adopts the facial geometry guidance as clues to accurately align and merge the frontal coarse HR face and prior information. Meanwhile, two-level discriminators are designed to capture both the global outline of a face image as well as detailed facial characteristics. The Coarse-D enforces the coarsely hallucinated faces to be upright and complete while the Fine-D focuses on the fine hallucinated ones for sharper details. Extensive experiments demonstrate that our VividGAN achieves photo-realistic frontal HR faces, reaching superior performance in downstream tasks, i.e., face recognition and expression classification, compared with other state-of-the-art methods.

This paper addresses the problem of cross-view image geo-localization, where the geographic location of a ground-level street-view query image is estimated by matching it against a large scale aerial map (e.g., a high-resolution satellite image). State-of-the-art deep-learning based methods tackle this problem as deep metric learning which aims to learn global feature representations of the scene seen by the two different views. Despite promising results are obtained by such deep metric learning methods, they, however, fail to exploit a crucial cue relevant for localization, namely, the spatial layout of local features. Moreover, little attention is paid to the obvious domain gap (between aerial view and ground view) in the context of cross-view localization. This paper proposes a novel Cross-View Feature Transport (CVFT) technique to explicitly establish cross-view domain transfer that facilitates feature alignment between ground and aerial images. Specifically, we implement the CVFT as network layers, which transports features from one domain to the other, leading to more meaningful feature similarity comparison. Our model is differentiable and can be learned end-to-end. Experiments on large-scale datasets have demonstrated that our method has remarkably boosted the state-of-the-art cross-view localization performance, e.g., on the CVUSA dataset, with significant improvements for top-1 recall from 40.79% to 61.43%, and for top-10 from 76.36% to 90.49%. We expect the key insight of the paper (i.e., explicitly handling domain difference via domain transport) will prove to be useful for other similar problems in computer vision as well.

Word-level sign language recognition (WSLR) is a fundamental task in sign language interpretation. It requires models to recognize isolated sign words from videos. However, annotating WSLR data needs expert knowledge, thus limiting WSLR dataset acquisition. On the contrary, there are abundant subtitled sign news videos on the internet. Since these videos have no word-level annotation and exhibit a large domain gap from isolated signs, they cannot be directly used for training WSLR models. We observe that despite the existence of a large domain gap, isolated and news signs share the same visual concepts, such as hand gestures and body movements. Motivated by this observation, we propose a novel method that learns domain-invariant visual concepts and fertilizes WSLR models by transferring knowledge of subtitled news sign to them. To this end, we extract news signs using a base WSLR model, and then design a classifier jointly trained on news and isolated signs to coarsely align these two domain features. In order to learn domain-invariant features within each class and suppress domain-specific features, our method further resorts to an external memory to store the class centroids of the aligned news signs. We then design a temporal attention based on the learnt descriptor to improve recognition performance. Experimental results on standard WSLR datasets show that our method outperforms previous state-of-the-art methods significantly. We also demonstrate the effectiveness of our method on automatically localizing signs from sign news, achieving 28.1 for [email protected].

Cross-view geo-localization is the problem of estimating the position and orientation (latitude, longitude and azimuth angle) of a camera at ground level given a large-scale database of geo-tagged aerial (eg., satellite) images. Existing approaches treat the task as a pure location estimation problem by learning discriminative feature descriptors, but neglect orientation alignment. It is well-recognized that knowing the orientation between ground and aerial images can significantly reduce matching ambiguity between these two views, especially when the ground-level images have a limited Field of View (FoV) instead of a full field-of-view panorama. Therefore, we design a Dynamic Similarity Matching network to estimate cross-view orientation alignment during localization. In particular, we address the cross-view domain gap by applying a polar transform to the aerial images to approximately align the images up to an unknown azimuth angle. Then, a two-stream convolutional network is used to learn deep features from the ground and polar-transformed aerial images. Finally, we obtain the orientation by computing the correlation between cross-view features, which also provides a more accurate measure of feature similarity, improving location recall. Experiments on standard datasets demonstrate that our method significantly improves state-of-the-art performance. Remarkably, we improve the top-1 location recall rate on the CVUSA dataset by a factor of 1.5x for panoramas with known orientation, by a factor of 3.3x for panoramas with unknown orientation, and by a factor of 6x for 180°-FoV images with unknown orientation.

Existing face hallucination methods based on convolutional neural networks (CNN) have achieved impressive performance on low-resolution (LR) faces in a normal illumination condition. However, their performance degrades dramatically when LR faces are captured in low or non-uniform illumination conditions. This paper proposes a Copy and Paste Generative Adversarial Network (CPGAN) to recover authentic high-resolution (HR) face images while compensating for low and non-uniform illumination. To this end, we develop two key components in our CPGAN: internal and external Copy and Paste nets (CPnets). Specifically, our internal CPnet exploits facial information residing in the input image to enhance facial details; while our external CPnet leverages an external HR face for illumination compensation. A new illumination compensation loss is thus developed to capture illumination from the external guided face image effectively. Furthermore, our method offsets illumination and upsamples facial details alternatively in a coarse-to-fine fashion, thus alleviating the correspondence ambiguity between LR inputs and external HR inputs. Extensive experiments demonstrate that our method manifests authentic HR face images in a uniform illumination condition and outperforms state-of-the-art methods qualitatively and quantitatively.

Compared with laborious pixel-wise dense labeling, it is much easier to label data by scribbles, which only costs 1~2 seconds to label one image. However, using scribble labels to learn salient object detection has not been explored. In this paper, we propose a weakly-supervised salient object detection model to learn saliency from such annotations. In doing so, we first relabel an existing large-scale salient object detection dataset with scribbles, namely S-DUTS dataset. Since object structure and detail information is not identified by scribbles, directly training with scribble labels will lead to saliency maps of poor boundary localization. To mitigate this problem, we propose an auxiliary edge detection task to localize object edges explicitly, and a gated structure-aware loss to place constraints on the scope of structure to be recovered. Moreover, we design a scribble boosting scheme to iteratively consolidate our scribble annotations, which are then employed as supervision to learn high-quality saliency maps. As existing saliency evaluation metrics neglect to measure structure alignment of the predictions, the saliency map ranking may not comply with human perception. We present a new metric, termed saliency structure measure, as a complementary metric to evaluate sharpness of the prediction. Extensive experiments on six benchmark datasets demonstrate that our method not only outperforms existing weakly-supervised/unsupervised methods, but also is on par with several fully-supervised state-of-the-art models (Our code and data is publicly available at: https://github.com/JingZhang617/Scribble_Saliency).

We propose a closed-loop, multi-instance control algorithm for visually guided reaching based on novel learning principles. A control Lyapunov function methodology is used to design a reaching action for a complex multi-instance task in the case where full state information (poses of all potential reaching points) is available. The proposed algorithm uses monocular vision and manipulator joint angles as the input to a deep convolution neural network to predict the value of the control Lyapunov function (cLf) and corresponding velocity control. The resulting network output is used in real-time as visual control for the grasping task with the multi-instance capability emerging naturally from the design of the control Lyapunov function. We demonstrate the proposed algorithm grasping mugs (textureless and symmetric objects) on a table-top from an over-the-shoulder monocular RGB camera. The manipulator dynamically converges to the best-suited target among multiple identical instances from any random initial pose within the workspace. The system trained with only simulated data is able to achieve 90.3% grasp success rate in the real-world experiments with up to 85Hz closed-loop control on one GTX 1080Ti GPU and significantly outperforms a Pose-Based-Visual-Servo (PBVS) grasping system adapted from a state-of-the-art single shot RGB 6D pose estimation algorithm. A key contribution of the paper is the inclusion of a first-order differential constraint associated with the cLf as a regularisation term during learning, and we provide evidence that this leads to more robust and reliable reaching/grasping performance than vanilla regression on general control inputs.

Sign language translation (SLT) aims to interpret sign video sequences into text-based natural language sentences. Sign videos consist of continuous sequences of sign gestures with no clear boundaries in between. Existing SLT models usually represent sign visual features in a frame-wise manner so as to avoid needing to explicitly segmenting the videos into isolated signs. However, these methods neglect the temporal information of signs and lead to substantial ambiguity in translation. In this paper, we explore the temporal semantic structures of sign videos to learn more discriminative features. To this end, we first present a novel sign video segment representation which takes into account multiple temporal granularities, thus alleviating the need for accurate video segmentation. Taking advantage of the proposed segment representation, we develop a novel hierarchical sign video feature learning method via a temporal semantic pyramid network, called TSPNet. Specifically, TSPNet introduces an inter-scale attention to evaluate and enhance local semantic consistency of sign segments and an intra-scale attention to resolve semantic ambiguity by using non-local video context. Experiments show that our TSPNet outperforms the state-of-the-art with significant improvements on the BLEU score (from 9.58 to 13.41) and ROUGE score (from 31.80 to 34.96) on the largest commonly used SLT dataset. Our implementation is available at https://github.com/verashira/TSPNet.

Vision-based sign language recognition aims at helping the deaf people to communicate with others. However, most existing sign language datasets are limited to a small number of words. Due to the limited vocabulary size, models learned from those datasets cannot be applied in practice. In this paper, we introduce a new large-scale Word-Level American Sign Language (WLASL) video dataset, containing more than 2000 words performed by over 100 signers. This dataset will be made publicly available to the research community. To our knowledge,it is by far the largest public ASL dataset to facilitate word-level sign recognition research. Based on this new large-scale dataset, we are able to experiment with several deep learning methods for word-level sign recognition and evaluate their performances in large scale scenarios. Specifically we implement and compare two different models,i.e., (i) holistic visual appearance based approach, and (ii) 2D human pose based approach. Both models are valuable baselines that will benefit the community for method benchmarking. Moreover, we also propose a novel pose-based temporal graph convolution networks (Pose-TGCN) that model spatial and temporal dependencies in human pose trajectories simultaneously, which has further boosted the performance of the pose-based method. Our results show that pose-based and appearance-based models achieve comparable performances up to 62.63% at top-10 accuracy on 2,000 words/glosses, demonstrating the validity and challenges of our dataset. Our dataset and baseline deep models are available at https://dxli94.github.io/WLASL/.

Given a tiny face image, existing face hallucination methods aim at super-resolving its high-resolution (HR) counterpart by learning a mapping from an exemplary dataset. Since a low-resolution (LR) input patch may correspond to many HR candidate patches, this ambiguity may lead to distorted HR facial details and wrong attributes such as gender reversal and rejuvenation. An LR input contains low-frequency facial components of its HR version while its residual face image, defined as the difference between the HR ground-truth and interpolated LR images, contains the missing high-frequency facial details. We demonstrate that supplementing residual images or feature maps with additional facial attribute information can significantly reduce the ambiguity in face super-resolution. To explore this idea, we develop an attribute-embedded upsampling network, which consists of an upsampling network and a discriminative network. The upsampling network is composed of an autoencoder with skip-connections, which incorporates facial attribute vectors into the residual features of LR inputs at the bottleneck of the autoencoder, and deconvolutional layers used for upsampling. The discriminative network is designed to examine whether super-resolved faces contain the desired attributes or not and then its loss is used for updating the upsampling network. In this manner, we can super-resolve tiny (16×16 pixels) unaligned face images with a large upscaling factor of 8× while reducing the uncertainty of one-to-many mappings remarkably. By conducting extensive evaluations on a large-scale dataset, we demonstrate that our method achieves superior face hallucination results and outperforms the state-of-the-art.

In popular TV programs (such as CSI), a very low-resolution face image of a person, who is not even looking at the camera in many cases, is digitally super-resolved to a degree that suddenly the person's identity is made visible and recognizable. Of course, we suspect that this is merely a cinematographic special effect and such a magical transformation of a single image is not technically possible. Or, is it? In this paper, we push the boundaries of super-resolving (hallucinating to be more accurate) a tiny, non-frontal face image to understand how much of this is possible by leveraging the availability of large datasets and deep networks. To this end, we introduce a novel Transformative Adversarial Neural Network (TANN) to jointly frontalize very-low resolution (i.e., 16 × 16 pixels) out-of-plane rotated face images (including profile views) and aggressively super-resolve them (8×), regardless of their original poses and without using any 3D information. TANN is composed of two components: a transformative upsampling network which embodies encoding, spatial transformation and deconvolutional layers, and a discriminative network that enforces the generated high-resolution frontal faces to lie on the same manifold as real frontal face images. We evaluate our method on a large set of synthesized non-frontal face images to assess its reconstruction performance. Extensive experiments demonstrate that TANN generates both qualitatively and quantitatively superior results achieving over 4 dB improvement over the state-of-the-art.

Portrait relighting aims to render a face image under different lighting conditions. Existing methods do not explicitly consider some challenging lighting effects such as specular and shadow, and thus may fail in handling extreme lighting conditions. In this paper, we propose a novel framework that explicitly models multiple reflectance channels for single image portrait relighting, including the facial albedo, geometry as well as two lighting effects, i.e. , specular and shadow. These channels are finally composed to generate the relit results via deep neural networks. Current datasets do not support learning such multiple reflectance channel modeling. Therefore, we present a large-scale dataset with the ground-truths of the channels, enabling us to train the deep neural networks in a supervised manner. Furthermore, we develop a novel module named Lighting guided Feature Modulation (LFM). In contrast to existing methods which simply incorporate the given lighting in the bottleneck of a network, LFM fuses the lighting by layer-wise feature modulation to deliver more convincing results. Extensive experiments demonstrate that our proposed method achieves better results and is able to generate challenging lighting effects.

Event-based cameras can measure intensity changes (called ‘events’) with microsecond accuracy under high-speed motion and challenging lighting conditions. With the active pixel sensor (APS), the event camera allows simultaneous output of the intensity frames. However, the output images are captured at a relatively low frame-rate and often suffer from motion blur. A blurry image can be regarded as the integral of a sequence of latent images, while the events indicate the changes between the latent images. Therefore, we are able to model the blur-generation process by associating event data to a latent image. In this paper, we propose a simple and effective approach, the Event-based Double Integral (EDI) model, to reconstruct a high frame-rate, sharp video from a single blurry frame and its event data. The video generation is based on solving a simple non-convex optimization problem in a single scalar variable. Experimental results on both synthetic and real images demonstrate the superiority of our EDI model and optimization method in comparison to the state-of-the-art.

Despite the fact that Second Order Similarity (SOS) has been used with significant success in tasks such as graph matching and clustering, it has not been exploited for learning local descriptors. In this work, we explore the potential of \sos in the field of descriptor learning by building upon the intuition that a positive pair of matching points should exhibit similar distances with respect to other points in the embedding space. Thus, we propose a novel regularization term, named Second Order Similarity Regularization (SOSR), that follows this principle. By incorporating SOSR into training, our learned descriptor achieves state-of-the-art performance on several challenging benchmarks containing distinct tasks ranging from local patch retrieval to structure from motion. Furthermore, by designing a von Mises-Fischer distribution based evaluation method, we link the utilization of the descriptor space to the matching performance, thus demonstrating the effectiveness of our proposed SOSR. Extensive experimental results, empirical evidence, and in-depth analysis are provided, indicating that SOSR can significantly boost the matching performance of the learned descriptor.

In this paper, we develop a new deep network to explicitly address these inherent differences between ground and aerial views. We observe there exist some approximate domain correspondences between ground and aerial images. Specifically, pixels lying on the same azimuth direction in an aerial image approximately correspond to a vertical image column in the ground view image. Thus, we propose a two-step approach to exploit this prior knowledge. The first step is to apply a regular polar transform to warp an aerial image such that its domain is closer to that of a ground-view panorama. Note that polar transform as a pure geometric transformation is agnostic to scene content, hence cannot bring the two domains into full alignment. Then, we add a subsequent spatial-attention mechanism which further brings corresponding deep features closer in the embedding space. To improve the robustness of feature representation, we introduce a feature aggregation strategy via learning multiple spatial embeddings. By the above two-step approach, we achieve more discriminative deep representations, facilitating cross-view Geo-localization more accurate. Our experiments on standard benchmark datasets show significant performance boosting, achieving more than doubled recall rate compared with the previous state of the art.

Recovering a photorealistic face from an artistic portrait is a challenging task since crucial facial details are often distorted or completely lost in artistic compositions. To handle this loss, we propose an Attribute-guided Face Recovery from Portraits (AFRP) that utilizes a Face Recovery Network (FRN) and a Discriminative Network (DN). FRN consists of an autoencoder with residual block-embedded skip-connections and incorporates facial attribute vectors into the feature maps of input portraits at the bottleneck of the autoencoder. DN has multiple convolutional and fully-connected layers, and its role is to enforce FRN to generate authentic face images with corresponding facial attributes dictated by the input attribute vectors. For the preservation of identities, we impose the recovered and ground-truth faces to share similar visual features. Specifically, DN determines whether the recovered image looks like a real face and checks if the facial attributes extracted from the recovered image are consistent with given attributes. Our method can recover photorealistic identity-preserving faces with desired attributes from unseen stylized portraits, artistic paintings, and hand-drawn sketches. On large-scale synthesized and sketch datasets, we demonstrate that our face recovery method achieves state-of-the-art results.

Estimating scene depth from a single image can be widely applied to understand 3D environments due to the easy access of the images captured by consumer-level cameras. Previous works exploit conditional random fields (CRFs) to estimate image depth, where neighboring pixels (superpixels) with similar appearances are constrained to share the same depth. However, the depth may vary significantly in the slanted surface, thus leading to severe estimation errors. In order to eliminate those errors, we propose a superpixel-based normal guided scale invariant deep convolutional field by encouraging the neighboring superpixels with similar appearance to lie on the same 3D plane of the scene. In doing so, a depth-normal multitask CNN is introduced to produce the superpixel-wise depth and surface normal predictions simultaneously. To correct the errors of the roughly estimated superpiexl-wise depth, we develop a normal guided scale invariant CRF (NGSI-CRF). NGSI-CRF consists of a scale invariant unary potential that is able to measure the relative depth between superpixels as well as the absolute depth of superpixels, and a normal guided pairwise potential that constrains spatial relationships between superpixels in accordance with the 3D layout of the scene. In other words, the normal guided pairwise potential is designed to smooth the depth prediction without deteriorating the 3D structure of the depth prediction. The superpixel-wise depth maps estimated by NGSI-CRF will be fed into a pixel-wise refinement module to produce a smooth fine-grained depth prediction. Furthermore, we derive a closed-form solution for the maximum a posteriori (MAP) inference of NGSI-CRF. Thus, our proposed network can be efficiently trained in an end-to-end manner. We conduct our experiments on various datasets, such as NYU-D2, KITTI, and Make 3D. As demonstrated in the experimental results, our method achieves superior performance in both indoor and outdoor scenes.

Given a tiny face image, existing face hallucination methods aim at super-resolving its high-resolution (HR) counterpart by learning a mapping from an exemplar dataset. Since a low-resolution (LR) input patch may correspond to many HR candidate patches, this ambiguity may lead to distorted HR facial details and wrong attributes such as gender reversal. An LR input contains low-frequency facial components of its HR version while its residual face image, defined as the difference between the HR ground-truth and interpolated LR images, contains the missing high-frequency facial details. We demonstrate that supplementing residual images or feature maps with additional facial attribute information can significantly reduce the ambiguity in face super-resolution. To explore this idea, we develop an attribute-embedded upsampling network, which consists of an upsampling network and a discriminative network. The upsampling network is composed of an autoencoder with skip-connections, which incorporates facial attribute vectors into the residual features of LR inputs at the bottleneck of the autoencoder and deconvolutional layers used for upsampling. The discriminative network is designed to examine whether super-resolved faces contain the desired attributes or not and then its loss is used for updating the upsampling network. In this manner, we can super-resolve tiny (16×16 pixels) unaligned face images with a large upscaling factor of 8× while reducing the uncertainty of one-to-many mappings remarkably. By conducting extensive evaluations on a large-scale dataset, we demonstrate that our method achieves superior face hallucination results and outperforms the state-of-the-art.

Estimating the disparity and normal direction of one pixel simultaneously, instead of only disparity, also known as 3D label methods, can achieve much higher subpixel accuracy in the stereo matching problem. However, it is extremely difficult to assign an appropriate 3D label to each pixel from the continuous label space R3 while maintaining global consistency because of the infinite parameter space. In this paper, we propose a novel algorithm called PatchMatch-based superpixel cut to assign 3D labels of an image more accurately. In order to achieve robust and precise stereo matching between local windows, we develop a bilayer matching cost, where a bottom-up scheme is exploited to design the two layers. The bottom layer is employed to measure the similarity between small square patches locally by exploiting a pretrained convolutional neural network, and then, the top layer is developed to assemble the local matching costs in large irregular windows induced by the tangent planes of object surfaces. To optimize the spatial smoothness of local assignments, we propose a novel strategy to update 3D labels. In the procedure of optimization, both segmentation information and random refinement of PatchMatch are exploited to update candidate 3D label set for each pixel with high probability of achieving lower loss. Since pairwise energy of general candidate label sets violates the submodular property of graph cut, we propose a novel multilayer superpixel structure to group candidate label sets into candidate assignments, which thereby can be efficiently fused by α-expansion graph cut. Extensive experiments demonstrate that our method can achieve higher subpixel accuracy in different data sets, and currently ranks first on the new challenging Middlebury 3.0 benchmark among all the existing methods.

We tackle the challenge of constructing 64 pixels for each individual pixel of a thumbnail face image. We show that such an aggressive super-resolution objective can be attained by taking advantage of the global context and making the best use of the prior information portrayed by the image class. Our input image is so small (e.g., pixels) that it can be considered as a patch of itself. Thus, conventional patch-matching-based super-resolution solutions are unsuitable. In order to enhance the resolution while enforcing the global context, we incorporate a pixel-wise appearance similarity objective into a deconvolutional neural network, which allows efficient learning of mappings between low-resolution input images and their high-resolution counterparts in the training data set. Furthermore, the deconvolutional network blends the learned high-resolution constituent parts in an authentic manner, where the face structure is naturally imposed and the global context is preserved. To account for the possible artifacts in upsampled feature maps, we employ a sub-network composed of additional convolutional layers. During training, we use roughly aligned images (only eye locations), yet demonstrate that our network has the capacity to super-resolve face images regardless of pose and facial expression variations. This significantly reduces the requirement of precisely face alignments in the data set. Owing to the network topology we apply, our method is robust to translational misalignments. In addition, our method is able to upsample rotational unaligned faces with data augmentation. Our extensive experimental analysis manifests that our method achieves more appealing and superior results than the state of the art.

Conventional face hallucination methods rely heavily on accurate alignment of low-resolution (LR) faces before upsampling them. Misalignment often leads to deficient results and unnatural artifacts for large upscaling factors. However, due to the diverse range of poses and different facial expressions, aligning an LR input image, in particular when it is tiny, is severely difficult. To overcome this challenge, here we present an end-to-end transformative discriminative neural network (TDN) devised for super-resolving unaligned and very small face images with an extreme upscaling factor of 8. Our method employs an upsampling network where we embed spatial transformation layers to allow local receptive fields to line-up with similar spatial supports. Furthermore, we incorporate a class-specific loss in our objective through a successive discriminative network to improve the alignment and upsampling performance with semantic information. Extensive experiments on large face datasets show that the proposed method significantly outperforms the state-of-the-art.

Most of the conventional face hallucination methods assume the input image is sufficiently large and aligned, and all require the input image to be noise-free. Their performance degrades drastically if the input image is tiny, unaligned, and contaminated by noise. In this paper, we introduce a novel transformative discriminative autoencoder to 8X super-resolve unaligned noisy and tiny (16X16) low-resolution face images. In contrast to encoder-decoder based autoencoders, our method uses decoder-encoder-decoder networks. We first employ a transformative discriminative decoder network to upsample and denoise simultaneously. Then we use a transformative encoder network to project the intermediate HR faces to aligned and noise-free LR faces. Finally, we use the second decoder to generate hallucinated HR images. Our extensive evaluations on a very large face dataset show that our method achieves superior hallucination results and outperforms the state-of-the-art by a large margin of 1.82dB PSNR.

How to build an accurate and reliable appearance model to improve the performance is a crucial problem in object tracking. Since the multi-view learning can lead to more accurate and robust representation of the object, in this paper, we propose a novel tracking method via multi-view learning framework by using multiple support vector machines (SVM). The multi-view SVMs tracking method is constructed based on multiple views of features and a novel combination strategy. To realize a comprehensive representation, we select three different types of features, i.e., gray scale value, histogram of oriented gradients (HOG), and local binary pattern (LBP), to train the corresponding SVMs. These features represent the object from the perspectives of description, detection, and recognition, respectively . In order to realize the combination of the SVMs under the multi-view learning framework, we present a novel collaborative strategy with entropy criterion, which is acquired by the confidence distribution of the candidate samples. In addition, to learn the changes of the object and the scenario, we propose a novel update scheme based on subspace evolution strategy. The new scheme can control the model update adaptively and help to address the occlusion problems . We conduct our approach on several public video sequences and the experimental results demonstrate that our method is robust and accurate, and can achieve the state-of-the-art tracking performance.

In this paper, we address the problem of estimating a latent sharp image from a single spatially variant blurred image. Non-uniform deblurring methods based on projective motion path models formulate the blur as a linear combination of homographic projections of a clear image. But they are computationally expensive and require large memory due to the calculation and storage of a large number of the projections. Patch-wise non-uniform deblurring algorithms have been proposed to estimate each kernel locally by a uniform deblurring algorithm, which does not require to calculate and store the projections. The key issues of these methods are the accuracy of kernel estimation and the identification of erroneous kernels. To perform accurate kernel estimation, we employ the total variation (TV) regularization to recover a latent image, in which the edges are better enhanced and the ringing artifacts are reduced, rather than Tikhonov regularization that previous algorithms adopt. Thus blur kernels can be estimated more accurately from the latent image and estimated in a closed form while previous methods cannot estimate kernels in closed forms. To identify the erroneous kernels, we develop a novel metric, which is able to measure the similarity between the neighboring kernels. After replacing the erroneous kernels with the well-estimated ones, a clear image is obtained. The experiments show that our approach can achieve better results on the real-world blurry images while using less computation and memory.