Unsupervised video Object-Centric Learning (OCL) is promising as it enables object-level scene representation and understanding as we humans do. Mainstream video OCL methods adopt a recurrent architecture: An aggregator aggregates current video frame into object features, termed slots, under some queries; A transitioner transits current slots to queries for the next frame. This is an effective architecture but all existing implementations both (i1) neglect to incorporate next frame features, the most informative source for query prediction, and (i2) fail to learn transition dynamics, the knowledge essential for query prediction. To address these issues, we propose Random Slot-Feature pair for learning Query prediction (RandSF.Q): (t1) We design a new transitioner to incorporate both slots and features, which provides more information for query prediction; (t2) We train the transitioner to predict queries from slot-feature pairs randomly sampled from available recurrences, which drives it to learn transition dynamics. Experiments on scene representation demonstrate that our method surpass existing video OCL methods significantly, e.g., up to 10 points on object discovery, setting new state-of-the-art. Such superiority also benefits downstream tasks like scene understanding.
论文检索
输入标题、作者或关键词，从 100,903 篇学术成果中精准定位
Infrared and Visible Image Fusion (IVIF) produces enhanced images by fusing complementary visual information. However, most existing methods generate fixed outputs and cannot flexibly adapt to user-specific requirements. Recent text-guided approaches offer partial control but are limited to global or semantic levels, lacking instance-level control. This limitation arises from two challenges: first, the lack of datasets that directly link textual instructions with corresponding spatial annotations, and second, the use of coarse cross-modal alignment methods that struggle to precisely match textual instructions with visual features. To overcome these challenges, we propose ControlFuse, a controllable IVIF framework enabling multi-granularity fusion across global, semantic, and instance levels, guided by user instructions. First, we construct an automated multi-granularity dataset that provides explicit textual-mask correspondences at these three levels. Second, inspired by manifold geometry, we design a Multimodal Feature Interaction Module (MFIM) comprising Feature Manifold Converter (FMC) and Curvature-Guided Interaction (CGI). FMC projects textual and visual features into a unified manifold space, while CGI leverages manifold curvature as a geometric cue to refine cross-modal alignment. Extensive experiments validate ControlFuse, outperforming state-of-the-art methods in robustness and flexibility.
Cheating Stereo Matching in Full-Scale: Physical Adversarial Attack Against Binocular Depth Estimation in Autonomous Driving
PDF ↗Though deep neural models adopted to realize the perception of autonomous driving have proven vulnerable to adversarial examples, known attacks often leverage 2D patches and target mostly monocular perception. Therefore, the effectiveness of Physical Adversarial Examples (PAEs) on stereo-based binocular depth estimation remains largely unexplored. To this end, we propose the first texture-enabled physical adversarial attack against stereo matching models in the context of autonomous driving. Our method employs a 3D PAE with global camouflage texture rather than a local 2D patch-based one, ensuring both visual consistency and attack effectiveness across different viewpoints of stereo cameras. To cope with the disparity effect of these cameras, we also propose a new 3D stereo matching rendering module that allows the PAE to be aligned with real-world positions and headings in binocular vision. We further propose a novel merging attack that seamlessly blends the target into the environment through fine-grained PAE optimization. It has significantly enhanced stealth and lethality upon existing hiding attacks that fail to get seamlessly merged into the background. Extensive evaluations show that our PAEs can successfully fool the stereo models into producing erroneous depth information.
Beyond Predictive Resampling: Learning Input-Agnostic Downsampling for Efficient Aligned Vision Recognition
PDF ↗Images are typically sampled on a uniform grid,despite their non-uniform information distribution—some regions are rich in content while others are not. The mismatch leads to inefficient computation allocation in deep learning models. To address this, recent studies have proposed predictive downsampling methodsthat adaptively downsample images based on predicted per-pixel importance, allocating more pixels to informative areas. However,these methods require high-resolution processing to accurately estimate importance, which undermines their efficiency:the prediction itself must process the full-resolution image,consuming most of the computational budget. This high-resolution importance prediction is necessary because each input may differ significantly in structure and content. In this paper, we take a different approach and introduce a learn-to-downsample paradigmtailored for aligned vision recognition tasks, such as face recognition and palmprint recognition, where input alignment ensures consistent spatial structure across images. This alignment ensures structural consistency across images, allowing a shared, input-agnostic downsampling template applicable to all inputs. Furthermore, instead of relying on implicit importance maps, we introduce a flow-based representation that explicitly models the spatial warping from the original image to the downsampled version. The flow representation is not only more efficient but also more controllable: we regularize the flow using its Jacobian determinant to precisely control the sampling density and coverage,enabling interpretable and tunable sampling patterns. Extensive experiments on two aligned recognition tasks, face and palmprint recognition, demonstrate that our method substantially reduces computational cost with minimal accuracy degradation, achieving a significantly better performance-efficiency trade-off than existing predictive downsampling methods.
Learning procedural-aware video representations is a key step towards building agents that can reason about and execute complex tasks. Existing methods typically address this problem by aligning visual content with textual descriptions at the task and step levels to inject procedural semantics into video representations. However, due to their high level of abstraction, "task" and "step" descriptions fail to form a robust alignment with the concrete, observable details in visual data. To address this, we introduce "states", i.e., textual snapshots of object configurations, as a visually-grounded semantic layer that anchors abstract procedures to what a model can actually see. We formalize this insight in a novel Task-Step-State (TSS) framework, where tasks are achieved via steps that drive transitions between observable states. To enforce this structure, we propose a progressive pre-training strategy that unfolds the TSS hierarchy, forcing the model to first ground representations in states before associating them with steps and, ultimately, high-level tasks. Extensive experiments on the COIN and CrossTask datasets show that our method outperforms baseline models on multiple downstream tasks, including task recognition, step recognition, and next step prediction. Ablation studies show that introducing state supervision is a key driver of performance gains across all tasks. Additionally, our progressive pretraining strategy proves more effective than standard joint training, as it better enforces the intended hierarchical structure.
Facial Dynamics in Video: Instruction Tuning for Improved Facial Expression Perception and Contextual Awareness
PDF ↗Facial expression captioning has found widespread application across various domains. Recently, the emergence of video Multimodal Large Language Models (MLLMs) has shown promise in general video understanding tasks. However, describing facial expressions within videos poses two major challenges for these models: (1) the lack of adequate datasets and benchmarks, and (2) the limited visual token capacity of video MLLMs. To address these issues, this paper introduces a new instruction-following dataset tailored for dynamic facial expression caption. The dataset comprises 5,033 high-quality video clips annotated manually, containing over 700,000 tokens. Its purpose is to improve the capability of video MLLMs to discern subtle facial nuances. Furthermore, we propose FaceTrack-MM, which leverages a limited number of tokens to encode the main character’s face. This model demonstrates superior performance in tracking faces and focusing on the facial expressions of the main characters, even in intricate multiperson scenarios. Additionally, we introduce a novel evaluation metric combining event extraction, relation classification, and the longest common subsequence (LCS) algorithm to assess the content consistency and temporal sequence consistency of generated text. Moreover, we present FECBench, a benchmark designed to assess the performance of existing video MLLMs in this specific task.
Multi-view 3D object detection plays a vital role in autonomous driving systems due to its ability to perceive complex scenes accurately. However, real-world driving data often exhibits a long-tailed distribution, causing significant drops in detection accuracy for rare categories in existing methods. To mitigate this issue, we propose CLIPDet3D, a novel vision-language collaborative framework for multi-view 3D object detection. First, to tackle the difficulty of capturing the semantic information of rare categories, a Vision-Language Collaborative Learning strategy is proposed to incorporate class-level semantic priors from CLIP. Second, a Depth Feature Contrastive Distillation module is designed to overcome the large depth estimation error for rare categories by aligning depth features between a teacher and a student network. Furthermore, to alleviate the difficulty in focusing on regions of rare categories, a Dual-Stream Prompt Attention mechanism is devised to inject learnable prompts and compute attention along both horizontal and vertical BEV directions. Evaluations on the nuScenes dataset demonstrate that CLIPDet3D achieves state-of-the-art accuracy while maintaining efficient inference.
Unified Representation Causal Prompt Distillation for Re-Inference-Free Lifelong Person Re-Identification
PDF ↗Lifelong person re-identification (LReID) aims to retrieve the target person from sequentially collected data. Due to significant domain gaps between datasets and the continuous increase of training data from different scenarios, weak inter-domain generalization and catastrophic forgetting issues have remained major challenges for LReID. To tackle these issues, a novel LReID method called Unified Representation Causal Prompt Distillation (URCPD) is proposed. Specifically, to reduce domain gaps among different scene datasets and improve model inter-domain generalization capability, a Feature Decoupling Style Transfer module (FDST) is proposed to map new features into a unified feature space. Furthermore, to reduce the accumulated forgetting of old knowledge during the training stage, a Causal Prompt Distillation module (CPD) is introduced. This module eliminates the re-inference process for distillation and embeds memory prompts to combat catastrophic forgetting. Extensive experiments on five classic LReID seen datasets and seven unseen datasets demonstrate that our method significantly outperforms state-of-the-art methods.
While current state-of-the-art Remote Sensing Change Detection (RSCD) methods can achieve impressive results on individual datasets, they become unreliable in unseen environments and imaging conditions, with performance metrics declining by as much as 60% to 80%. Simultaneously, variable environments and complex imaging conditions are the main characteristics of remote sensing data, calling for generalizable RSCD methods. To address this issue, we propose a novel RSCD method capable of domain generalization—CDDGNet. This method is based on causal decoupling theory, which progressively decouples invariant change features from variable domain features to extract generalizable characteristics. This enables a network trained on a single domain to accurately identify change regions in other domains. Specifically, firstly, the Causal Feature Adaptation Module is proposed to preliminarily decouple and simplify feature information during the encoding process by using wavelet transformation and feature energy spectralization methods. Secondly, the Causal Feature Fusion Module is presented to fully decouple features and aggregate significant change features during the decoding process through frequency domain processing and feature re-attention mechanisms. Thirdly, the Decoupling Effect Loss Function is proposed to optimize the process by evaluating the effectiveness of causal decoupling. Extensive experiments have shown that our model significantly outperforms existing methods across multiple groups of generalization tasks with varying levels of difficulty.
A dexterous hand capable of generalizable grasping objects is fundamental for the development of general-purpose embodied AI. However, previous methods focus narrowly on low-level grasp stability metrics, neglecting affordance-aware positioning and human-like poses which are crucial for downstream manipulation. To address these limitations, we propose AffordDex, a novel framework with two-stage training that learns a universal grasping policy with an inherent understanding of both motion priors and object affordances. In the first stage, a trajectory imitator is pre-trained on a large corpus of human hand motions to instill a strong prior for natural movement. In the second stage, a residual module is trained to adapt these general human-like motions to specific object instances. This refinement is critically guided by two components: our Negative Affordance-aware Segmentation (NAA) module, which identifies functionally inappropriate contact regions, and a privileged teacher-student distillation process that ensures the final vision-based policy is highly successful. Extensive experiments demonstrate that AffordDex not only achieves universal dexterous grasping but also remains remarkably human-like in posture and functionally appropriate in contact location. As a result, AffordDex significantly outperforms state-of-the-art baselines across seen objects, unseen instances, and even entirely novel categories.
Concept Bottleneck Models (CBMs) enhance interpretability by introducing a layer of human-understandable concepts between inputs and predictions. While recent methods automate concept generation using Large Language Models (LLMs) and Vision-Language Models (VLMs), they still face three fundamental challenges: poor visual grounding, concept redundancy, and the absence of principled metrics to balance predictive accuracy and concept compactness. We introduce PS-CBM, a Partially Shared CBM framework that addresses these limitations through three core components: (1) a multimodal concept generator that integrates LLM-derived semantics with exemplar-based visual cues; (2) a Partially Shared Concept Strategy that merges concepts based on activation patterns to balance specificity and compactness; and (3) Concept-Efficient Accuracy (CEA), a post-hoc metric that jointly captures both predictive accuracy and concept compactness. Extensive experiments on eleven diverse datasets show that PS-CBM consistently outperforms state-of-the-art CBMs, improving classification accuracy by 1.0%–7.4% and CEA by 2.0%–9.5%, while requiring significantly fewer concepts. These results underscore PS-CBM’s effectiveness in achieving both high accuracy and strong interpretability.
Tracking Any Point (TAP) is a foundational task in computer vision with broad applicability. The state-of-the-art self-supervised TAP method leverages a global matching transformer and contrastive random walks to learn point correspondences. However, its dense all-pairs attention and correlation volume computation tend to introduce irrelevant features and produce less informative training signals, degrading both learning efficiency and tracking accuracy. To address these limitations, we introduce LEAP-Track, a self-supervised TAP approach that computes the attention matrices and correlation volume over adaptively selected sparse pairs. It consists of two core designs: (1) Curriculum-based Sparse Attention (CSA), which dynamically focuses on the most relevant keys, promoting the learning of discriminative features; and (2) Progressive k-NN Transition (PkT), which reformulates the contrastive random walk to operate on an increasingly sparse k-NN affinity graph to reinforce the learning of the most informative correspondences. By integrating the above two designs into a two-stage training paradigm, LEAP-Track is shown both theoretically and empirically to effectively boost learning efficiency, achieving superior tracking accuracy over existing self-supervised TAP methods.
Real-time 3D object detection from point clouds is essential for dynamic scene understanding in applications such as augmented reality, robotics, and navigation. We introduce a novel Spatial-prioritized and Rank-aware 3D object detection (SR3D) framework for indoor point clouds, to bridge the gap between how detectors are trained and how they are evaluated. This gap stems from the lack of spatial reliability and ranking awareness during training, which conflicts with the ranking-based prediction selection used at inference. Such a training-inference gap hampers the model’s ability to learn representations aligned with inference-time behavior. To address the limitation, SR3D consists of two components tailored to the spatial nature of point clouds during training: a novel spatial-prioritized optimal transport assignment that dynamically emphasizes well-located and spatially reliable samples, and a rank-aware adaptive self-distillation scheme that adaptively injects ranking perception via a self-distillation paradigm. Extensive experiments on ScanNet V2 and SUN RGB-D show that SR3D effectively bridges the training-inference gap and significantly outperforms prior methods in accuracy while maintaining real-time speed.
MAISI-v2: Accelerated 3D High-Resolution Medical Image Synthesis with Rectified Flow and Region-specific Contrastive Loss
PDF ↗Medical image synthesis is an important topic for both clinical and research applications. Recently, diffusion models have become a leading approach in this area. Despite their strengths, many existing methods struggle with (1) limited generalizability, only working for specific body regions or voxel spacings, (2) slow inference, which is a common issue for diffusion models, and (3) weak alignment with input conditions, which is a critical issue for medical imaging. MAISI, a previously proposed framework, addresses generalizability issues but still suffers from slow inference and limited condition consistency. In this work, we present MAISI-v2, the first accelerated 3D medical image synthesis framework that integrates rectified flow to enable fast and high-quality generation. To further enhance condition fidelity, we introduce a novel region-specific contrastive loss to improve sensitivity to the region of interest. Our experiments show that MAISI-v2 can achieve state-of-the-art image quality with 33× acceleration for latent diffusion models. We also conducted a downstream segmentation experiment to show that the synthetic images can be used for data augmentation. We release our code, training details, model weights, and a GUI demo to facilitate reproducibility and promote further development within the community.
Aiming to estimate the full extent of partially occluded objects, amodal segmentation is a critical capability for visual intelligence. Existing methods suffer from limitations in efficiency and precision, due to their reliance on auxiliary information or two-stage architectures. Furthermore, they lack generalizability, failing to meet practical requirements. To overcome these challenges, we proposed a new paradigm, CondDiff-AMO, that interprets amodal segmentation as a denoising problem by leveraging diffusion models. Methodologically, the designed novel framework consists of three key innovations to adapt the task characteristics and unlocks the diffusion models’ potential in amodal segmentation, including a masking strategy in the forward process, an adaptive transformer for conditional feature extraction, and visual-guided sampling. In the forward process, progressive masking strategy converts ground-truth masks to visible masks, simulating amodal segmentation process to enhance reasoning regarding occluded areas. For architectural design, a pyramid network with feature refinement extracts adaptive and representative conditional priors, improving the guidance in the denoising process of diffusion models. As for the sampling stage, a visible mask is incorporated with an ensemble strategy, restricting the prediction on occluded part. Experiments were conducted on five well-known datasets under supervised and zero-shot learning, with the results confirming that CondDiff-AMO outperforms state-of-the-art methods.
Diffusion Distillation with Direct Preference Optimization for Efficient 3D LiDAR Scene Completion
PDF ↗The slow sampling speed of diffusion models hinders their application in 3D LiDAR scene completion. To address this, we propose Distillation-DPO, a novel framework that accelerates sampling through score distillation while simultaneously enhancing generation quality via preference alignment. Distillation-DPO follows a three-step procedure. First, the student model generates paired completion scenes with different initial noises. Second, using LiDAR scene evaluation metrics as preference, we construct winning and losing sample pairs. Third, as our core innovation, Distillation-DPO optimizes the student model by exploiting the difference in score functions between the teacher and student models on the paired completion scenes. This operation performs variational score distillation of the student model but simultaneously encourages the distilled student to prefer the winning samples over the losing ones. Extensive experiments demonstrate that Distillation-DPO achieves higher-quality scene completion than state-of-the-art diffusion models, while accelerating sampling by over 5-fold. To our knowledge, our work is the first to integrate the preference learning principle of DPO into the distillation of diffusion models, offering a new framework of preference-aligned distillation.
Simulating Distribution Dynamics: Liquid Temporal Feature Evolution for Single-Domain Generalized Object Detection
PDF ↗In this paper, we focus on Single-Domain Generalized Object Detection (Single-DGOD), aiming to transfer a detector trained on one source domain to multiple unknown domains. Existing methods for Single-DGOD typically rely on discrete data augmentation or static perturbation methods to expand data diversity, thereby mitigating the lack of access to target domain data. However, in real-world scenarios such as changes in weather or lighting conditions, domain shifts often occur continuously and gradually. Discrete augmentations and static perturbations fail to effectively capture the dynamic variation of feature distributions, thereby limiting the model's ability to perceive fine-grained cross-domain differences. To this end, we propose a new method, i.e., Liquid Temporal Feature Evolution, which simulates the progressive evolution of features from the source domain to simulated latent distributions by incorporating temporal modeling and liquid neural network–driven parameter adjustment. Specifically, we introduce controllable Gaussian noise injection and multi-scale Gaussian blurring to simulate initial feature perturbations, followed by temporal modeling and a liquid parameter adjustment mechanism to generate adaptive modulation parameters, enabling a smooth and continuous adaptation across domains. By capturing progressive cross-domain feature evolution and dynamically regulating adaptation paths, our method bridges the source-unknown domain distribution gap, significantly boosting generalization and robustness to unseen shifts. Significant performance improvements on the Diverse Weather dataset and Real-to-Art benchmark demonstrate the superiority of our method.
Disentangling for Transfer: Boosting Limited Modalities via Information-Theoretic Regularization and Cross-Modal Reconstruction
PDF ↗Missing critical modalities in medical imaging poses significant challenges for AI-driven diagnostic systems, particularly in scenarios where limited modalities must suffice for downstream tasks. Existing approaches often fail to fully leverage privileged features available only at training or address the information gap between privileged and limited modalities, resulting in suboptimal performance. To address this, we propose a unified, dual-stage Disentanglement-AligNmenT framEwork (DANTE), which uses InformationTheoretic Regularization and Cross-Modal Reconstruction to decompose full-modality information into alignable and privileged-exclusive components. In the first stage, a self-supervised pre-training strategy based on cross-modal reconstruction acts as a proxy task to implicitly incentivize disentangled representations. In the second stage, we present an information-theoretic regularization to explicitly maximize the transfer of privileged knowledge through two novel modules: (1) a Mutual Alignment Module that employs multilevel bidirectional alignment between limited-modality features and alignable features, enhancing cross-modal representation consistency; (2) a Privileged Compaction Module that restricts the privileged-exclusive information flow, promoting the integration of task-relevant content into alignable representations. Experimental results on three challenging medical datasets demonstrate that DANTE achieves state-of-the-art performance, demonstrating its effectiveness in leveraging privileged guidance under modality scarcity, and exhibits broad applicability across diverse medical imaging scenarios.
Pose-agnostic Anomaly Detection (PAD) aims to detect anomalies when the poses of query images are unknown and differ from those in the training set. Therefore, accurately estimating the camera poses for the query images in the test set is critical for this task. Existing query-specific framework methods require re-optimizing a new set of parameters for each query image, limiting their generalization and increasing computational burden. To overcome these limitations, we propose a novel method, Relative Pose Estimation for Pose-agnostic Anomaly Detection (RPE-PAD), which enhances both generalization and efficiency with a query-independent framework. Specifically, we propose a Random View Synthesis Scheme (RVSS) that generates new poses by adding Gaussian perturbations to the original poses, then renders the corresponding views to augment the dataset. To estimate the relative camera pose between two input images, we introduce an Iterative Relative Pose Refinement Network (IRPRN), which incorporates a hierarchical coarse-to-fine refinement strategy. Furthermore, we employ a Multi-Pair Training Strategy (MPTS) to train the proposed IRPRN, leveraging multiple image pairs to expand the relative pose transformation space during training. Extensive experiments demonstrate that our method achieves robust anomaly detection performance while significantly improving inference efficiency.
D²-VPR: A Parameter-efficient Visual-foundation-model-based Visual Place Recognition Method via Knowledge Distillation and Deformable Aggregation
PDF ↗Visual Place Recognition (VPR) aims to determine the geographic location of a query image by retrieving its most visually similar counterpart from a geo-tagged reference database. Recently, the emergence of the powerful visual foundation model, DINOv2, trained in a self-supervised manner on massive datasets, has significantly improved VPR performance. This improvement stems from DINOv2’s exceptional feature generalization capabilities but is often accompanied by increased model complexity and computational overhead that impede deployment on resource-constrained devices. To address this challenge, we propose D2-VPR, a Distillation- and Deformable-based framework that retains the strong feature extraction capabilities of visual foundation models while significantly reducing model parameters and achieving a more favorable performance-efficiency trade-off. Specifically, first, we employ a two-stage training strategy that integrates knowledge distillation and fine-tuning. Additionally, we introduce a Distillation Recovery Module (DRM) to better align the feature spaces between the teacher and student models, thereby minimizing knowledge transfer losses to the greatest extent possible. Second, we design a Top-Down-attention-based Deformable Aggregator (TDDA) that leverages global semantic features to dynamically and adaptively adjust the Regions of Interest (ROI) used for aggregation, thereby improving adaptability to irregular structures. Extensive experiments demonstrate that our method achieves competitive performance compared to state-of-the-art approaches. Meanwhile, it reduces the parameter count by approximately 64.2% (compared to CricaVPR).