论文检索

Wei Dong, Han Zhou, Junwei Lin, Jun Chen

Real-world dark images commonly exhibit not only low visibility and contrast but also complex noise and blur, posing significant restoration challenges. Existing methods often rely on paired data or fail to model dynamic illumination and blur characteristics, leading to poor generalization. To tackle this, we propose a generative framework based on visual autoregressive (VAR) modeling, guided by perceptual priors from the vision-language model (VLM). Specifically, to supply informative conditioning cues for VAR models, we deploy an adaptive curve estimation scheme to modulate the diverse illumination based on VLM-derived visibility scores. In addition, we integrate dynamic and spatial-frequency-aware Rotary Positional Encodings (SF-RoPE) into VAR to enhance its ability to model structures degraded by blur. Furthermore, we propose a recursive phase-domain modulation strategy that mitigates blur-induced artifacts in the phase domain via bounded iterative refinement guided by VLM-assessed blur scores. Our framework is fully unsupervised and achieves state-of-the-art performance on benchmark datasets.

Linfeng Dong, Yuchen Yang, Hao Wu, Wei Wang, Yuenan Hou, Zhihang Zhong, Xiao Sun

We introduce RacketVision, a novel dataset and benchmark for advancing computer vision in sports analytics, covering table tennis, tennis, and badminton. The dataset is the first to provide large-scale, fine-grained annotations for racket pose alongside traditional ball positions, enabling research into complex human-object interactions. It is designed to tackle three interconnected tasks: fine-grained ball tracking, articulated racket pose estimation, and predictive ball trajectory forecasting. Our evaluation of established baselines reveals a critical insight for multi-modal fusion: while naively concatenating racket pose features degrades performance, a Cross-Attention mechanism is essential to unlock their value, leading to trajectory prediction results that surpass strong unimodal baselines. RacketVision provides a versatile resource and a strong starting point for future research in dynamic object tracking, conditional motion forecasting, and multi-modal analysis in sports.

Jiahua Dong, Xudong Wang, Wenqi Liang, Zongyan Han, Meng Cao, Duzhen Zhang, Hanbin Zhao, Zhi Han, Salman Khan, Fahad Shahbaz Khan

Customized text-to-video generation (CTVG) has recently witnessed great progress in generating tailored videos from user-specific text. However, most CTVG methods assume that personalized concepts remain static and do not expand incrementally over time. Additionally, they struggle with forgetting and concept neglect when continuously learning new concepts, including subjects and motions. To resolve the above challenges, we develop a novel Continual Customized Video Diffusion (CCVD) model, which can continuously learn new concepts to generate videos across various text-to-video generation tasks by tackling forgetting and concept neglect. To address catastrophic forgetting, we introduce a concept-specific attribute retention module and a task-aware concept aggregation strategy. They can capture the unique characteristics and identities of old concepts during training, while combining all subject and motion adapters of old concepts based on their relevance during testing. Besides, to tackle concept neglect, we develop a controllable conditional synthesis to enhance regional features and align video contexts with user conditions, by incorporating layer-specific region attention-guided noise estimation. Extensive experimental comparisons demonstrate that our CCVD outperforms existing CTVG models.

Yijie Ding, Yang Liu, Haobo Jiang, Jianmin Zheng

Reconstructing precise CAD modeling sequences from point clouds remains a challenging task, especially for objects with complex geometry and topology. In this paper, by formulating the CAD sequence reconstruction as a Markov decision process, we introduce ReACT, a novel Reward-informed Autoregressive decision Cad Transformer architecture for robust CAD sequence prediction. Beyond previous imitation-only approaches, our key innovation is to frame the CAD Transformer under a reinforcement learning paradigm and thereby integrate reward-inspired heuristic learning into our architecture. This allows ReACT to effectively leverage shape-aware long-term reward feedback to guide the inference of (nearly) optimal CAD commands. Specifically, conditioned on past tokens, comprising the historical CAD states, sketch-extrude commands (i.e., actions) and associated geometric rewards, ReACT autoregressively outputs the most promising CAD commands in a causal manner. In particular, we develop a novel scaffold-aware CAD state representation that integrates global point-command features with an incrementally constructed surface point scaffold, enabling fine-grained geometric reasoning for subsequent reconstruction prediction. Moreover, an effective local barrel points-guided dense reward function is designed to jointly evaluate surface fidelity and command efficiency for reliable reward guidance. Extensive evaluations on the DeepCAD and Fusion360 benchmarks demonstrate that ReACT can achieve superior CAD reconstruction quality, even for objects with complex shapes.

Rui Ding, Zhaonian Kuang, Zongwei Zhou, Meng Yang, Xinhu Zheng, Gang Hua

Multi-view 3D detection with bird’s eye view (BEV) is crucial for autonomous driving and robotics, but its robustness in real-world is limited as it struggles to predict accurate depth values. A mainstream solution, cross-modal distillation, transfers depth information from LiDAR to camera models but also unintentionally transfers depth-irrelevant information (e.g. LiDAR density). To mitigate this issue, we propose RayD3D, which transfers crucial depth knowledge along the ray: a line projecting from the camera to true location of an object. It is based on the fundamental imaging principle that predicted location of this object can only vary along this ray, which is finally determined by predicted depth value. Therefore, distilling along the ray enables more effective depth information transfer. More specifically, we design two ray-based distillation modules. Ray-based Contrastive Distillation (RCD) incorporates contrastive learning into distillation by sampling along the ray to learn how LiDAR accurately locates objects. Ray-based Weighted Distillation (RWD) adaptively adjusts distillation weight based on the ray to minimize the interference of depth-irrelevant information in LiDAR. For validation, we widely apply RayD3D into three representative types of BEV-based models, including BEVDet, BEVDepth4D, and BEVFormer. Our method is trained on clean NuScenes, and tested on both clean NuScenes and RoboBEV with a variety types of data corruptions. Our method significantly improves the robustness of all the three base models in all scenarios without increasing inference costs, and achieves the best when compared to recently released multi-view and distillation models.

Lei Ding, Xibing Zuo, Haitao Guo, Jun Lu, Zhihui Gong, Xuanguang Liu, Jicang Lu

Unsupervised Change Detection (UCD) in Very High Resolution (VHR) Remote Sensing (RS) images remains to be a difficult challenge due to the inherent spatio-temporal complexity within data. Inspired by recent advancements in Visual Foundation Models (VFMs) and Contrastive Learning (CL), this research aims to develop CL methodologies to translate implicit knowledge in VFM into change representations, thus eliminating the need for explicit supervision. To this end, we introduce a Semantic-to-Change (S2C) learning framework for UCD in VHR RS images. Differently from existing CL methodologies that typically focus on learning multi-temporal similarities, we introduce a novel triplet learning strategy that explicitly models temporal differences, which are crucial to the CD task. Furthermore, random spatial and spectral perturbations are introduced during training to enhance robustness to temporal noise. In addition, a grid sparsity regularization is defined to suppress insignificant changes, and an IoU-matching algorithm is developed to refine the CD results. Experiments on three benchmark CD datasets demonstrate that the proposed S2C learning framework achieves significant improvements in accuracy, surpassing current state-of-the-art by over 31%, 9% and 23%, respectively. It also demonstrates robustness and sample efficiency, suitable for training and adaptation of various VFMs or backbone neural networks.

Ankit Dhiman, Srinath R, Jaswanth Reddy, Lokesh R Boregowda, Venkatesh Babu Radhakrishnan

3D Gaussian Splatting (3DGS) and Neural Radiance Fields (NeRF) have advanced novel-view synthesis. Recent methods extend multi-view 2D segmentation to 3D, enabling instance/semantic segmentation for better scene understanding. A key challenge is the inconsistency of 2D instance labels across views, leading to poor 3D predictions. Existing methods use a two-stage approach in which some rely on contrastive learning with hyperparameter-sensitive clustering, while others preprocess labels for consistency. We propose a unified framework that merges these steps, reducing training time and improving performance by introducing a learnable feature embedding for segmentation in Gaussian primitives. This embedding is then efficiently decoded into instance labels through a novel "Embedding-to-Label" process, effectively integrating the optimization. While this unified framework offers substantial benefits, we observed artifacts at the object boundaries. To address the object boundary issues, we propose hard-mining samples along these boundaries. However, directly applying hard mining to the feature embeddings proved unstable. Therefore, we apply a linear layer to the rasterized feature embeddings before calculating the triplet loss, which stabilizes training and significantly improves performance. Our method outperforms baselines qualitatively and quantitatively on the ScanNet, Replica3D, and Messy-Rooms datasets.

Zeyu Deng, Lihui Wang, Xi Tao, Qijian Chen, Ying Cao, XuLinHu, Yingfeng Ou

The inherently low signal-to-noise ratio (SNR) in diffusion-weighted (DW) imaging fundamentally impedes precise tissue microstructure characterization, rendering effective noise suppression a persistent challenge. Existing denoising methods frequently suffer from over-smoothing or distortion of microstructure information when handling spatially correlated or severe noise. To address these limitations, we propose UP2-MAE fusion model, a self-supervised DWI denoising method based on Uncertainty-Propelled Physics and Masked Auto-Encoder (MAE) fusion. This framework integrates two complementary branches: one leverages MAE to suppress noise through local context modeling, while the other constructs uncorrelated noisy pairs using diffusion tensor imaging (DTI) physics and denoises them via a Noise2Noise approach, which can preserve texture details by exploiting directional relationships across diffusion encoding directions. To fully integrate the strengths of both branches, an uncertainty-propelled fusion strategy based on maximum likelihood estimation is proposed to derive the final denoised output. In addition, to further promote the performance, uncertainty-guided reconstruction and consistency loss are presented. Evaluations against state-of-the-art denoising methods on both simulated and acquired DW datasets confirm the efficacy of our approach.

Yuxin Deng, Botian Wang, Kaining Zhang, Hao Zhang, Jiayi Ma

Multi-modal image matching is a fundamental task in multi-view and multi-modal image processing. Its key challenge lies in extracting features that remain consistent despite drastic appearance variations across modalities. However, the learning of the feature is hindered by the scarcity and the inaccurate alignment of existing multi-modal datasets. To address this, we propose a knowledge distillation framework termed SGPFeat that transfers rich prior knowledge from large-scale unimodal tasks to enhance multi-modal representation learning. Specifically, semantic priors from a vision foundation model guide the feature extractor to identify shared semantic structures across modalities, enabling better generalization under large appearance gaps. In parallel, geometric priors derived from accurately aligned visible-light datasets improve detection precision on noisy aligned multi-modal pairs. Furthermore, we introduce a Heterogeneous Feature Aggregation (HFA) module to facilitate effective distillation and feature representation. Extensive experiments demonstrate that semantic and geometric priors bring significant improvement for our SGPFeat across diverse multi-modal image matching benchmarks.

Yang Deng, Zhanke Wang, Jiahao Wu, Jie Liang, Jingui Ma, Yang Hu, Ronggang Wang

Reconstructing 3D scenes from multi-view image sequences remains a significant challenge in practical applications. While recent advances in 3D Gaussian Splatting have enabled high-quality rendering, existing methods rely heavily on pixel-level L1 loss, which misaligns with human perception, leading to a lack of high-frequency details and the emergence of artifacts. Additionally, the position gradient-based densification strategy often results in under-densified Gaussian primitives, thereby degrading rendering quality. To address these challenges, we propose Pano-GS, a perception-aware Gaussian optimization framework. Specifically, we introduce a gradient consistency-constrained loss to capture high-frequency details, mitigating the inherent shortcomings of traditional L1 loss and enhancing reconstruction fidelity. In addition, we use a multi-criteria densification strategy to reduce the sole reliance on average position gradients. Extensive experiments demonstrate that Pano-GS achieves state-of-the-art performance, confirming its effectiveness and robust generalization across diverse real-world scenes.

Ruxi Deng, Wenxuan Bao, Tianxin Wei, Jingrui He

Pretrained vision-language models exhibit strong zero-shot classification capabilities, but their predictions degrade significantly under common image corruptions. To improve robustness, many test-time adaptation (TTA) methods adopt positive data augmentation (PDA), which generates multiple views of each test sample to reduce prediction variance. However, these methods suffer from two key limitations. First, it introduces considerable computational overhead due to the large number of augmentations required per image. Second, it fails to mitigate prediction bias, where the model tends to predict certain classes disproportionately under corruption, as PDA operates on corrupted inputs and typically does not remove the corruption itself. To address these challenges, we propose Panda, a novel TTA method based on negative data augmentation (NDA). Unlike positive augmentations that preserve object semantics, Panda generates negative augmentations by disrupting semantic content. It divides images into patches and randomly assembles them from a shared patch pool. These negatively augmented images retain corruption-specific features while discarding object-relevant signals. We then subtract the mean feature of these negative samples from the original image feature, effectively suppressing corruption-related components while preserving class-relevant information. This mitigates prediction bias under distribution shifts. Importantly, Panda allows augmentation to be shared across samples within a batch, resulting in minimal computational overhead. Panda can be seamlessly integrated into existing test-time adaptation frameworks and substantially improve their robustness. Our experiments indicate that Panda delivers superior performance compared to PDA methods, and a wide range of TTA methods exhibit significantly enhanced performance when integrated with Panda.

Dasith De Silva Edirimuni, Ajmal Saeed Mian

Most 3D scene generation methods are limited to only generating object bounding box parameters while newer diffusion methods also generate class labels and latent features. Using object size or latent feature, they then retrieve objects from a predefined database. For complex scenes of varied, multi-categorical objects, diffusion-based latents cannot be effectively decoded by current autoencoders into the correct point cloud objects which agree with target classes. We introduce a Class-Partitioned Vector Quantized Variational Autoencoder (CPVQ-VAE) that is trained to effectively decode object latent features, by employing a pioneering class-partitioned codebook where codevectors are labeled by class. To address the problem of codebook collapse, we propose a class-aware running average update which reinitializes dead codevectors within each partition. During inference, object features and class labels, both generated by a Latent-space Flow Matching Model (LFMM) designed specifically for scene generation, are consumed by the CPVQ-VAE. The CPVQ-VAE's class-aware inverse look-up then maps generated latents to codebook entries that are decoded to class-specific point cloud shapes. Thereby, we achieve pure point cloud generation without relying on an external objects database for retrieval. Extensive experiments reveal that our method reliably recovers plausible point cloud scenes, with up to 70.4% and 72.3% reduction in Chamfer and Point2Mesh errors on complex living room scenes.

Edoardo De Matteis, Matteo Migliarini, Alessio Sampieri, Indro Spinelli, Fabio Galasso

We introduce Human Motion Unlearning and motivate it through the concrete task of preventing violent 3D motion synthesis, an important safety requirement given that popular text-to-motion datasets (HumanML3D and Motion-X) contain from 7% to 15% violent sequences spanning both atomic gestures (e.g., a single punch) and highly compositional actions (e.g., loading and swinging a leg to kick). By focusing on violence unlearning, we demonstrate how removing a challenging, multifaceted concept can serve as a proxy for the broader capability of motion "forgetting." To enable systematic evaluation of Human Motion Unlearning, we establish the first motion unlearning benchmark by automatically filtering HumanML3D and Motion-X datasets to create distinct forget sets (violent motions) and retain sets (safe motions). We introduce evaluation metrics tailored to sequential unlearning, measuring both suppression efficacy and the preservation of realism and smooth transitions. We adapt two state-of-the-art, training-free image unlearning methods (UCE and RECE) to leading text-to-motion architectures (MoMask and BAMM), and propose Latent Code Replacement (LCR), a novel, training-free approach that identifies violent codes in a discrete codebook representation and substitutes them with safe alternatives. Our experiments show that unlearning violent motions is indeed feasible and that acting on latent codes strikes the best trade-off between violence suppression and preserving overall motion quality. This work establishes a foundation for advancing safe motion synthesis across diverse applications.

Nastaran Darabi, Devashri Naik, Sina Tayebati, Dinithi Jayasuriya, Ranganath Krishnan, Amit Ranjan Trivedi

Large Language Models (LLMs) and Vision-Language Models (VLMs) remain highly vulnerable to adversarial attacks despite widespread adoption. Existing defenses typically require retraining, rely on heuristics, or fail under adaptive and out-of-distribution (OOD) conditions. We introduce EigenShield, a principled, inference-time, architecture-agnostic defense that leverages Random Matrix Theory (RMT) to suppress adversarial noise in high-dimensional embeddings. EigenShield uses spiked covariance modeling and a Robustness-based Nonconformity Score (RbNS) with quantile thresholding to isolate and preserve causal eigenvectors, filtering out adversarial components without model access or adversarial training. We develop a theoretical framework establishing conditions for asymptotic noise suppression and demonstrate effectiveness in both unimodal and multimodal settings. Empirically, EigenShield consistently improves robustness across threat models, reducing attack success rates (ASR) by up to 48% over state-of-the-art defenses, including adversarial training, UNIGUARD, CIDER, and input transformations. On jailbreak attacks, EigenShield lowers LLM ASR by up to 92.9% relative to undefended models. Under multimodal adversarial attacks, it reduces VLM ASR by up to 76.5%. Against adaptive attacks on LLMs, it achieves ASR reductions of up to 77.7%. In OOD settings, EigenShield maintains strong performance, reducing ASR by up to 88.4% for LLMs and 80.4% for VLMs.

Thinh Dao, Khoa D Doan, Kok-Seng Wong

Deep Neural Networks (DNNs) are shown to be vulnerable to backdoor poisoning attacks, with most research focusing on digital triggers that consist of artificial patterns added to test-time inputs to induce targeted misclassification. Physical triggers, which are natural objects embedded in real-world scenes, offer a promising alternative for attackers as they can activate backdoors in real-time without digital manipulation. However, existing physical backdoor attacks are dirty-label, meaning that attackers must change the labels of poisoned inputs to the target label. The inconsistency between image content and label exposes the attack to human inspection, reducing its stealthiness in real-world settings. To address this limitation, we introduce Clean-Label Physical Backdoor Attack (CLPBA), a new paradigm of physical backdoor attack that does not require label manipulation and trigger injection at the training stage. Instead, the attacker injects imperceptible perturbations into a small number of target class samples to backdoor a model. By framing the attack as a Dataset Distillation problem, we develop three CLPBA variants, namely Parameter Matching, Gradient Matching, and Feature Matching, that craft effective poisons under both linear probing and full-finetuning training settings. In hard scenarios that require backdoor generalizability in the physical world, CLPBA is shown to even surpass Dirty-label attack baselines. We demonstrate the effectiveness of CLPBA via extensive experiments on two collected physical backdoor datasets for facial recognition and animal classification.

Jisheng Dang, Delin Deng, Bimei Wang, Jingze Wu, Hui Zhang, Haijiang Li, Jingmei Jiao, Dengyue Pan, Mangang Xie, Jizhao Liu

Despite significant advancements in point cloud analysis, reducing energy consumption and improving robustness remain understudied, largely due to the inherent limitations of Convolutional Neural Networks (CNNs). To address this, we take the cue from the primary visual cortex and propose a Dendritic-Connected Continuous-Coupled Neural Network (DC-CCNN), a novel Brain-Inspired Neural Network (BINN) architecture tailored for point cloud analysis. By leveraging the unique characteristics of point clouds, our design combines discrete and continuous encoding, replacing traditional Multilayer Perceptrons (MLPs) with more efficient and robust BINNs. Our approach substantially improves the performance of Brain-Inspired Neural Networks on point analysis tasks and maintaining performance comparable to state-of-the-art methods. Furthermore, DC-CCNN exhibits enhanced robustness against various point cloud deformations and corruptions. Our experimental results demonstrate that DC-CCNN achieves competitive performance on benchmark datasets, making it a promising alternative to traditional deep learning methods for point cloud analysis. With its high efficiency and robustness, DC-CCNN has the potential for widespread adoption in 3D computer vision, robotics, and autonomous systems.

Chenxu Dang, Haiyan Liu, Jason Bao, Pei An, Xinyue Tang, An Pan, Jie Ma, Bingchuan Sun, Yan Wang

Semantic occupancy has emerged as a powerful representation in world models for its ability to capture rich spatial semantics. However, most existing occupancy world models rely on static and fixed embeddings or grids, which inherently limit the flexibility of perception. Moreover, their ``in-place classification" over grids exhibits a potential misalignment with the dynamic and continuous nature of real scenarios. In this paper, we propose SparseWorld, a novel 4D occupancy world model that is flexible, adaptive, and efficient, powered by sparse and dynamic queries. We propose a Range-Adaptive Perception module, in which learnable queries are modulated by the ego vehicle states and enriched with temporal-spatial associations to enable extended-range perception. To effectively capture the dynamics of the scene, we design a State-Conditioned Forecasting module, which replaces classification-based forecasting with regression-guided formulation, precisely aligning the dynamic queries with the continuity of the 4D environment. In addition, We specifically devise a Temporal-Aware Self-Scheduling training strategy to enable smooth and efficient training. Extensive experiments demonstrate that SparseWorld achieves state-of-the-art performance across perception, forecasting, and planning tasks. Comprehensive visualizations and ablation studies further validate the advantages of SparseWorld in terms of flexibility, adaptability, and efficiency.

Sisi Dai, Kai Xu

Despite significant progress in text-driven 4D human-object interaction (HOI) generation with supervised methods, the scalability remains limited by the scarcity of large-scale 4D HOI datasets. To overcome this, recent approaches attempt zero-shot 4D HOI generation with pre-trained image diffusion models. However, interaction cues are minimally distilled during the generation process, restricting their applicability across diverse scenarios. In this paper, we propose AnchorHOI, a novel framework that thoroughly exploits hybrid priors by incorporating video diffusion models beyond image diffusion models, advancing 4D HOI generation. Nevertheless, directly optimizing high-dimensional 4D HOI with such priors remains challenging, particularly for human pose and compositional motion. To address this challenge, AnchorHOI introduces an anchor-based prior distillation strategy, which constructs interaction-aware anchors and then leverages them to guide generation in a tractable two-step process. Specifically, two tailored anchors are designed for 4D HOI generation: anchor Neural Radiance Fields (NeRFs) for expressive interaction composition, and anchor keypoints for realistic motion synthesis. Extensive experiments demonstrate that AnchorHOI outperforms previous methods with superior diversity and generalization.

Longchen Dai, Zixuan Shen, Zhiheng Zhou, Peipeng Yu, Zhihua Xia

Face recognition systems store face templates for efficient matching. Once leaked, these templates pose a threat: inverting them can yield photorealistic surrogates that compromise privacy and enable impersonation. Although existing research has achieved relatively realistic face template inversion, the reconstructed facial images exhibit over-smoothed facial-part attributes (eyes, nose, mouth) and limited transferability. To address this problem, we present CLIP-FTI, a CLIP-driven fine-grained attribute conditioning framework for face template inversion. Our core idea is to use the CLIP model to obtain the semantic embeddings of facial features, in order to realize the reconstruction of specific facial feature attributes. Specifically, facial feature attribute embeddings extracted from CLIP are fused with the leaked template via a cross-modal feature interaction network and projected into the intermediate latent space of a pretrained Style- GAN. The StyleGAN generator then synthesizes face images with the same identity as the templates but with more finegrained facial feature attributes. Experiments across multiple face recognition backbones and datasets show that our reconstructions (i) achieve higher identification accuracy and attribute similarity, (ii) recover sharper component-level attribute semantics, and (iii) improve cross-model attack transferability compared to prior reconstruction attacks. To the best of our knowledge, ours is the first method to use additional information besides the face template attack to realize face template inversion and obtains SOTA results.

Linrui Dai, Rongzhao Zhang, Yongrui Yu, Xiaofan Zhang

The recently emerging conditional diffusion models seem promising for mitigating the labor and expenses in building large 3D medical imaging datasets. However, previous studies on 3D CT generation primarily focus on specific organs characterized by a local structure and fixed contrast and have yet to fully capitalize on the benefits of both semantic and textual conditions. In this paper, we present GuideGen, a controllable framework based on easily-acquired text prompts to generate anatomical masks and corresponding CT volumes for the entire torso—from chest to pelvis. Our approach includes three core components: a text-conditional semantic synthesizer for creating realistic full-torso anatomies; an anatomy-aware high-dynamic-range (HDR) autoencoder for high-fidelity feature extraction across varying intensity levels; and a latent feature generator that ensures alignment between CT images, anatomical semantics and input prompts. Combined, these components enable data synthesis for segmentation tasks from only textual instructions. To train and evaluate GuideGen, we compile a multi-modality cancer imaging dataset with paired CT and clinical descriptions from 12 public TCIA datasets and one private real-world dataset. Comprehensive evaluations across generation quality, cross-modality alignment, and data usability on multi-organ and tumor segmentation tasks demonstrate GuideGen's superiority over existing CT generation methods.