Point clouds, as a primary representation of 3D data, can be categorized into scene domain point clouds and object domain point clouds. Point cloud self-supervised learning (SSL) has become a mainstream paradigm for learning 3D representations. However, existing point cloud SSL primarily focuses on learning domain-specific 3D representations within a single domain, neglecting the complementary nature of cross-domain knowledge, which limits the learning of 3D representations. In this paper, we propose to learn a comprehensive Point cloud Mixture-of-Domain-Experts model (Point-MoDE) via a block-to-scene pre-training strategy. Specifically, We first propose a mixture-of-domain-expert model consisting of scene domain experts and multiple shared object domain experts. Furthermore, we propose a block-to-scene pretraining strategy, which leverages the features of point blocks in the object domain to regress their initial positions in the scene domain through object-level block mask reconstruction and scene-level block position regression. By integrating the complementary knowledge between object and scene, this strategy simultaneously facilitates the learning of both object-domain and scene-domain representations, leading to a more comprehensive 3D representation. Extensive experiments in downstream tasks demonstrate the superiority of our model.
论文检索
输入标题、作者或关键词，从 1,280 篇学术成果中精准定位
Since the collection of labeled gaze data is laborious and time-consuming, methods which can learn generalizable features by leveraging large-scale available unlabeled data are desirable. In recent years, we have witnessed the tremendous capabilities of diffusion models in generating images as well as their potential in feature representation learning. In this paper, we investigate whether they can acquire discriminative representations for gaze estimation via generative pre-training. To achieve this goal, we propose a self-supervised learning framework with diffusion models for gaze estimation, called GazeDiff. Specifically, we utilize a conditional diffusion model to generate target image with gaze direction specified by the reference image as the pre-training task. To facilitate the diffusion model to learn gaze related features as condition, we propose a disentangling feature learning strategy, which first learns appearance feature, head pose feature, and eye direction feature respectively, and then combines them as the conditional features. Extensive experiments demonstrate denoising diffusion models are also good general gaze feature learners.
Video Large Language Models (VideoLLMs) often require thousands of visual tokens to process long videos, leading to substantial computational costs, further exacerbated by visual token inefficiency. Existing token reduction and alternative video representation methods improve efficiency but often compromise comprehension abilities. In this work, we analyze the reasoning processes of VideoLLMs in multi-choice VideoQA task, identifying three reasoning stages—shallow, intermediate, and deep stages—that closely mimic human cognitive processing. Our analysis reveals specific inefficiencies at each stage: in shallow layers, VideoLLMs attempt to memorize all video details without prioritizing relevant content; in intermediate layers, models fail to re-examine uncertain content dynamically; and in deep layers, they continue processing video even when sufficiently confident. To bridge this gap, we propose DToMA, a training-free Dynamic Token MAnipulation method inspired by human adjustment mechanisms in three aspects: 1) Text-guided keyframe-aware reorganization to prioritize keyframes and reduce redundancy, 2) Uncertainty-based visual injection to revisit content dynamically, and 3) Early-exit pruning to halt visual tokens when confident. Experiments on 6 long video understanding benchmarks show that DToMA enhances both efficiency and comprehension, outperforming state-of-the-art methods and generalizing well across 3 VideoLLM architectures and sizes. Code is available at https://github.com/yuanrr/DToMA.
The hierarchical architecture has become a mainstream design paradigm for Vision Transformers (ViTs), with Patch Merging serving as the pivotal component that transforms a columnar architecture into a hierarchical one. Drawing inspiration from the brain's ability to integrate global and local information for comprehensive visual understanding, we propose Stepwise Patch Merging (SPM), which enhances the subsequent attention mechanism's ability to 'see' better. SPM consists of Multi-Scale Aggregation (MSA) and Guided Local Enhancement (GLE) striking a proper balance between long-range dependency modeling and local feature enhancement. Extensive experiments conducted on benchmark datasets, including ImageNet-1K, COCO, and ADE20K, demonstrate that SPM significantly improves the performance of various models, particularly in dense prediction tasks such as object detection and semantic segmentation. Meanwhile, experiments show that combining SPM with different backbones can further improve performance. The code has been released at https://github.com/Yonghao-Yu/StepwisePatchMerging.
Industrial video anomaly detection aims to perform real-time analysis of video streams from industrial production lines and provide anomaly alerts. Conventional video anomaly detection methods focus more on the overall image, as they aim to identify anomalies among multiple normal samples appearing simultaneously. However, industrial scenarios, where the primary focus is on a single type of product, require attention to local areas to capture fine-grained details and specific patterns. Directly applying conventional methods to industrial scenarios can result in an inability to focus on products moving along fixed trajectories, ineffective utilization of their equidistant periodicity, and greater susceptibility to lighting variations. To address these issues, we propose FreqNet, an encoder-decoder framework that learns frequency-domain features from videos to capture periodic and dynamic characteristics, enhancing the model's robustness. Specifically, a trajectory filter is proposed that takes advantage of the significant difference between moving objects and static backgrounds in the frequency domain by assigning higher weights to fixed moving trajectories. Moreover, a multi-feature fusion module is proposed, in which the frequency domain features of the video are first extracted to leverage the unique equidistant periodicity information of videos from industrial production lines. The extracted frequency domain features are subsequently fused with spatio-temporal features and contextual information is further integrated from the fused representation, effectively mitigating the impact of lighting variations on production lines. Extensive experiments on the benchmark IPAD dataset demonstrate the superiority of our proposed method over the state-of-the-art.
Video understanding seeks to enable machines to interpret visual content across three levels: action, event, and story. Existing models are limited in their ability to perform high-level long-term story understanding, due to (1) the oversimplified treatment of temporal information and (2) the training bias introduced by action/event-centric datasets. To address this, we introduce SCVBench, a novel benchmark for story-centric video understanding. SCVBench evaluates LVLMs through an event ordering task decomposed into sub-questions leading to a final question, quantitatively measuring historical dialogue exploration. We collected 1,253 final questions and 6,027 sub-question pairs from 925 videos, constructing continuous multi-turn dialogues. Experimental results show that while closed-source GPT-4o outperforms other models, most open-source LVLMs struggle with story-centric video understanding. Additionally, our StoryCoT model significantly surpasses open-source LVLMs on SCVBench. SCVBench aims to advance research by comprehensively analyzing LVLMs' temporal reasoning and comprehension capabilities. Code can be accessed at https://github.com/yuanrr/SCVBench.
In recent years, weakly supervised object detection (WSOD) has attracted much attention due to its low labeling cost. The success of recent WSOD models is often ascribed to the two-stage multi-class classification (MCC) task, i.e., multiple instance learning and online classification refinement. Despite achieving non-trivial progresses, these methods overlook potential classification ambiguities between these two MCC tasks and fail to leverage their unique strengths. In this work, we introduce a novel WSOD framework to ameliorate these two issues. For one thing, we propose a self-classification enhancement module that integrates intra-class binary classification (ICBC) to bridge the gap between the two distinct MCC tasks. The ICBC task enhances the network’s discrimination between positive and mis-located samples in a class-wise manner and forges a mutually reinforcing relationship with the MCC task. For another, we propose a self-classification correction algorithm during inference, which combines the results of both MCC tasks to effectively reduce the mis-classified predictions. Extensive experiments on the prevalent VOC 2007 & 2012 datasets demonstrate the superior performance of our framework.
With the rapid development of the internet, sharing photos through Social Network Platforms (SNPs) has become a new way for people to socialize, which poses serious threats to personal privacy. Recently, a thumbnail-preserving image privacy protection technique has emerged and garnered widespread attention. However, the existing schemes based on this technique often introduce noticeable noise into the protected image, resulting in poor visual quality. Motivated by the observation that a latent vector can be decoupled into the detail and contour components, in this paper, we propose HIPP, a thumbnail-preserving image privacy protection scheme that decouples the detail and contour information contained in the latent vector corresponding to the original image and reconstructs details by generation model. As a result, the generated protected image appears natural and has a thumbnail similar to the original one. Moreover, the protected images can be restored to versions that are indistinguishable from the original images. Experiments on CelebA, Helen, and LSUN datasets show that the SSIM between the restored and original images achieves 0.9899. Furthermore, compared to the previous works, HIPP achieves the lowest runtime and file expansion rate, with values of 0.07 seconds and 1.1046, respectively.
Class-agnostic counting enables enumerating arbitrary object classes beyond those seen during training. Recent studies attempted to exploit the potential of visual foundation models such as GroundingDINO. Despite the considerable progress, we observe certain shortcomings, including the limited diversity of visual prompts and suboptimal training regimen. To address these issues, we introduce VQCounter, which incorporates a visual prompt queue mechanism designed to enrich the diversity of visual prompts. A random modality switching strategy is proposed during training to strengthen both textual and visual modalities. Besides, in light of weak point supervision, a Voronoi diagram-based cost (VoronoiCost) is designed to improve Hungarian matching, leading to more stable and faster convergence. Building upon the Voronoi diagram, we also propose a novel set of more stringent evaluation metrics, which take point localization into account. Extensive experiments on the FSC-147 and CARPK datasets demonstrate that VQCounter achieves state-of-the-art performance in both zero-shot and few-shot settings, significantly outperforming existing methods across nearly all evaluations.
This paper tackles the challenge of anomaly image synthesis and segmentation to generate various anomaly images and their segmentation labels to mitigate the issue of data scarcity. Existing approaches employ the precise mask to guide the generation, relying on additional mask generators, leading to increased computational costs and limited anomaly diversity. Although a few works use coarse masks as the guidance to expand diversity, they lack effective generation of labels for synthetic images, thereby reducing their practicality. Therefore, our proposed method simultaneously generates anomaly images and their corresponding masks by utilizing coarse masks and anomaly categories. The framework utilizes attention maps from synthesis process as mask labels and employs two optimization modules to tackle drift challenges, which are mismatches between synthetic results and real situations. Our evaluation demonstrates that our method improves pixel-level AP by 1.3% and F1-MAX by 1.8% in anomaly detection tasks on the MVTec dataset. Additionally, its successful application in practical scenarios highlights its effectiveness, improving IoU by 37.2% and F-measure by 25.1% with the Floor Dirt dataset. The code is available at https://github.com/JJessicaYao/DriftRemover.
3D Single Object Tracking (SOT) is a fundamental task in computer vision and plays a critical role in applications like autonomous driving. However, existing algorithms often involve complex designs and multiple loss functions, making model training and deployment challenging. Furthermore, their reliance on fixed probability distribution assumptions (e.g., Laplacian or Gaussian) hinders their ability to adapt to diverse target characteristics such as varying sizes and motion patterns, ultimately affecting tracking precision and robustness. To address these issues, we propose BEVTrack, a simple yet effective motion-based tracking method. BEVTrack directly estimates object motion in Bird's-Eye View (BEV) using a single regression loss. To enhance accuracy for targets with diverse attributes, it learns adaptive likelihood functions tailored to individual targets, avoiding the limitations of fixed distribution assumptions in previous methods. This approach provides valuable priors for tracking and significantly boosts performance. Comprehensive experiments on KITTI, NuScenes, and Waymo Open Dataset demonstrate that BEVTrack achieves state-of-the-art results while operating at 200 FPS, enabling real-time applicability. The code will be released at https://github.com/xmm-prio/BEVTrack.
Do You Steal My Model? Signature Diffusion Embedded Dual-Verification Watermarking for Protecting Intellectual Property of Hyperspectral Image Classification Models
PDF ↗Due to the high cost of data collection and training, the well-performed hyperspectral image (HSI) classification models are of great value and vulnerable to piracy threat during transmission and use. Model watermarking is a promising technology for intellectual property (IP) protection of models. However, the existing model watermarking methods for RGB image classification models ignore the complexity of ground objects and high dimension of HSIs, which makes trigger samples easy to be detected and forged. To address this problem, we propose a signature diffusion embedded dual-verification watermarking method, which generates imperceptible trigger samples with explicit owner information to achieve dual verification of both model ownership and legality of trigger set. Specifically, the subpixel-space owner signature diffusion incorporated imperceptible trigger set generation method is proposed to manipulate owner signature incorporated to the abundance matrix of seeds via diffusion model in subpixel space, thus balancing the perceptual quality of trigger samples and signature extraction capability. To resist ownership confusion, dual-stamp ownership verification is proposed to query the suspicious model with trigger samples for ownership verification, and further extracts signature from trigger samples to guarantee their legality. Extensive experiments demonstrate the proposed method can effectively protect IP of HSI classification models.
DPMamba: Distillation Prompt Mamba for Multimodal Remote Sensing Image Classification with Missing Modalities
PDF ↗Multimodal remote sensing image classification (RSIC) has emerged as a key focus in Earth observation, driven by its capacity to extract complementary information from diverse sources. Existing methods struggle with modality absence caused by weather or equipment failures, leading to performance degradation. As a solution, knowledge distillation-based methods train student networks (SN) using a full-modality teacher, but they usually require training separate SN for each modality absence scenario, increasing complexity. To this end, we propose a unified Distillation Prompt Mamba (DPMamba) framework for multimodal RSIC with missing modalities. DPMamba leverages knowledge distillation in a shared text semantic space to optimize learnable prompts, transforming them from ``placeholder" to ``adaptation" states by enriching missing modality information with full-modality knowledge. To achieve this, we focus on two main aspects: first, we propose a new modality-aware Mamba for dynamically and hierarchically extracting cross-modality interactive features, providing richer, contextually relevant representations for backpropagation-based optimization of prompts; and second, we introduce a novel text-bridging distillation method to efficiently transfer full-modality knowledge, guiding the inclusion of missing modality information into prompts. Extensive evaluations demonstrate the effectiveness and robustness of the proposed DPMamba.
RAMer: Reconstruction-based Adversarial Model for Multi-party Multi-modal Multi-label Emotion Recognition
PDF ↗Conventional Multi-modal multi-label emotion recognition (MMER) assumes complete access to visual, textual, and acoustic modalities. However, real-world multi-party settings often violate this assumption, as non-speakers frequently lack acoustic and textual inputs, leading to a significant degradation in model performance. Existing approaches also tend to unify heterogeneous modalities into a single representation, overlooking each modality’s unique characteristics. To address these challenges, we propose RAMer (Reconstruction-based Adversarial Model for Emotion Recognition), which refines multi-modal representations by not only exploring modality commonality and specificity but crucially by leveraging reconstructed features, enhanced by contrastive learning, to overcome data incompleteness and enrich feature quality. RAMer also introduces a personality auxiliary task to complement missing modalities using modality-level attention, improving emotion reasoning. To further strengthen the model's ability to capture label and modality interdependency, we propose a stack shuffle strategy to enrich correlations between labels and modality-specific features. Experiments on three benchmarks, i.e., MEmoR, CMU-MOSEI, and M³ED, demonstrate that RAMer achieves state-of-the-art performance in dyadic and multi-party MMER scenarios.
The differences in imaging devices cause multimodal images to have modal differences and geometric distortions, complicating the matching task. Deep learning-based matching methods struggle with multimodal images due to the lack of large annotated multimodal datasets. To address these challenges, we propose XCP-Match based on cross-modality completion pre-training. XCP-Match has two phases. (1) Self-supervised cross-modality completion pre-training based on real multimodal image dataset. We develop a novel pre-training model to learn cross-modal semantic features. The pre-training uses masked image modeling method for cross-modality completion, and introduces an attention-weighted contrastive loss to emphasize matching in overlapping areas. (2) Supervised fine-tuning for multimodal image matching based on the augmented MegaDepth dataset. XCP-Match constructs a complete matching framework to overcome geometric distortions and achieve precise matching. Two-phase training encourages the model to learn deep cross-modal semantic information, improving adaptation to modal differences without needing large annotated datasets. Experiments demonstrate that XCP-Match outperforms existing algorithms on public datasets.
Accurate segmentation of lesions is crucial for disease diagnosis and treatment planning. However, blurring and low contrast in the imaging process can affect segmentation results. We have observed that noninvasive medical imaging shares considerable similarities with natural images under low light conditions and that nocturnal animals possess extremely strong night vision capabilities. Inspired by the dark vision of these nocturnal animals, we proposed a novel plug-and-play dark vision network (DVNet) to enhance the model's perception for low-contrast medical images. Specifically, by employing the wavelet transform, we decompose medical images into subbands of varying frequencies, mimicking the sensitivity of photoreceptor cells to different light intensities. To simulate the antagonistic receptive fields of horizontal cells and bipolar cells, we design a Mamba-Enhanced Fusion Module to achieve global information correlation and enhance contrast between lesions and surrounding healthy tissues. Extensive experiments demonstrate that the DVNet achieves SOTA performance in various medical image segmentation tasks.
AttentionDrag: Exploiting Latent Correlation Knowledge in Pre-trained Diffusion Models for Image Editing
PDF ↗Traditional point-based image editing methods rely on iterative latent optimization or geometric transformations, which are either inefficient in their processing or fail to capture the semantic relationships within the image. These methods often overlook the powerful yet underutilized image editing capabilities inherent in pre-trained diffusion models. In this work, we propose a novel one-step point-based image editing method, named \textbf{AttentionDrag}, which leverages the inherent latent knowledge and feature correlations within pre-trained diffusion models for image editing tasks. This framework enables semantic consistency and high-quality manipulation without the need for extensive re-optimization or retraining. Specifically, we reutilize the latent correlations knowledge learned by the self-attention mechanism in the U-Net module during the DDIM inversion process to automatically identify and adjust relevant image regions, ensuring semantic validity and consistency. Additionally, AttentionDrag adaptively generates masks to guide the editing process, enabling precise and context-aware modifications with friendly interaction. Our results demonstrate a performance that surpasses most state-of-the-art methods with significantly faster speeds, showing a more efficient and semantically coherent solution for point-based image editing tasks. Code is released at: https://github.com/GPlaying/AttentionDrag.
Compositional zero-shot learning (CZSL) aims to recognize novel compositions of attributes and objects learned from seen compositions. Previous works disentangle attributes and objects by extracting shared and exclusive parts between the image pair sharing the same attribute (object), as well as aligning them with pretrained word embeddings to improve unseen attribute-object recognition. Despite the significant achievements of existing efforts, they are hampered by three limitations: (1) The efficacy of disentanglement is compromised due to the influence of the background and the intricate entanglement of attributes with objects in the same parts. (2) Existing word embeddings fail to capture complex multimodal semantic information. (3) Overconfidence exhibited by existing models in seen compositions hinders their generalization to novel compositions. Being aware of these, we propose a novel framework named multimodal large language model (MLLM) embeddings and attribute smoothing guided disentanglement for CZSL. First, we leverage feature adaptive aggregation modules to mitigate the impact of background, and utilize learnable condition masks to capture multi-granularity features for disentanglement. Moreover, the last hidden states of MLLM are employed as word embeddings for their superior representation capabilities. Furthermore, we propose attribute smoothing with auxiliary attributes generated by the large language model (LLM) for seen compositions to address the overconfidence challenge. Extensive experiments demonstrate that our method achieves state-of-the-art performance on three challenging datasets. The supplementary material and source code will be available at https://github.com/xud-yan/Trident.
SCOUT: Semi-supervised Camouflaged Object Detection by Utilizing Text and Adaptive Data Selection
PDF ↗The difficulty of pixel-level annotation has significantly hindered the development of the Camouflaged Object Detection (COD) field. To save on annotation costs, previous works leverage the semi-supervised COD framework that relies on a small number of labeled data and a large volume of unlabeled data. We argue that there is still significant room for improvement in the effective utilization of unlabeled data. To this end, we introduce a Semi-supervised Camouflaged Object Detection by Utilizing Text and Adaptive Data Selection (SCOUT). It includes an Adaptive Data Augment and Selection (ADAS) module and a Text Fusion Module (TFM). The ADSA module selects valuable data for annotation through an adversarial augment and sampling strategy. The TFM module further leverages the selected valuable data by combining camouflage-related knowledge and text-visual interaction. To adapt to this work, we build a new dataset, namely RefTextCOD. Extensive experiments show that the proposed method surpasses previous semi-supervised methods in the COD field and achieves state-of-the-art performance. Our code will be released at https://github.com/Heartfirey/UCOD-DPL.
In large aperture imaging, the shallow depth of field (DoF) phenomenon requires capturing multiple images at different focal levels, allowing us to infer depth information using depth from focus (DFF) techniques. However, most previous works design convolutional neural networks from a time domain perspective, often leading to blurred fine details in depth estimation. In this work, we propose a frequency-aware deep DFF network (FAD) that couples multi-scale spatial domain local features with frequency domain global structural features. Our main innovations include two key points: First, we introduce a frequency domain feature extraction module that uses the Fourier transform to transfer latent focus features into the frequency domain. This module adaptively captures essential frequency information for focus changes through element-wise multiplication, enhancing fine details in depth results while preserving global structural integrity. Second, the time-frequency joint module of FAD improves the consistency of depth information in sparse texture regions and the continuity in transition areas from both local and global complementary perspectives. Comprehensive experiments demonstrate that our model achieves compelling generalization and state-of-the-art depth prediction across various datasets. Additionally, it can be quickly adapted to real-world applications as a pre-trained model.