AAAI 2026 论文列表

Accurate vessel segmentation in X-ray angiograms is crucial for numerous clinical applications. However, the scarcity of annotated data presents a significant challenge, which has driven the adoption of self-supervised learning (SSL) methods such as masked image modeling (MIM) to leverage large-scale unlabeled data for learning transferable representations. Unfortunately, conventional MIM often fails to capture vascular anatomy because of the severe class imbalance between vessel and background pixels, leading to weak vascular representations. To address this, we introduce Vascular anatomy-aware Masked Image Modeling (VasoMIM), a novel MIM framework tailored for X-ray angiograms that explicitly integrates anatomical knowledge into the pre-training process. Specifically, it comprises two complementary components: anatomy-guided masking strategy and anatomical consistency loss. The former preferentially masks vessel-containing patches to focus the model on reconstructing vessel-relevant regions. The latter enforces consistency in vascular semantics between the original and reconstructed images, thereby improving the discriminability of vascular representations. Empirically, VasoMIM achieves state-of-the-art performance across three datasets. These findings highlight its potential to facilitate X-ray angiogram analysis.

Source-free unsupervised domain adaptation (SF-UDA), which relies only on a pre-trained source model and unlabeled target data, has gained significant attention. Pseudo-labeling, valued for its simplicity and effectiveness, is a key approach in SF-UDA. However, existing methods neglect the consistency priors of anatomical features across samples, leading them fail to revise of high-confidence noise in structurally inconsistent regions, ultimately manifesting as significant discrepancies in pseudo-labeled samples especially in limited source data scenarios. Motivated by this insight, we propose a novel Geometric Correspondence Constrained (GCC) pseudo-labeling framework. GCC first stratifies pseudo-labeled samples into high/low-quality subsets. It then refines low-quality samples by leveraging the anatomical features inherent in high-quality samples while injecting Gaussian perturbation to perturb high-confidence noise towards the decision boundaries. This process effectively mitigates high-confidence noise disruptive effect and preserves critical prior anatomical knowledge, making it particularly powerful for scenarios with limited source data. Experiments on cross-domain fundus image datasets demonstrate that our method achieves state-of-the-art performance.

Capsule Network (CapsNet) has demonstrated significant potential in visual recognition by capturing spatial relationships and part-whole hierarchies for learning equivariant feature representations. However, existing CapsNet and variants often rely on a single high-level feature map, overlooking the rich complementary information provided by multi-scale features. Furthermore, conventional feature fusion strategies, such as addition and concatenation, struggle to reconcile multi-scale feature discrepancies, leading to suboptimal classification performance. To address these limitations, we propose the Multi-Scale Patchify Capsule Network (MSPCaps), a novel architecture that integrates multi-scale feature learning and efficient capsule routing. Specifically, MSPCaps consists of three key components: a Multi-Scale ResNet Backbone (MSRB), a Patchify Capsule Layer (PatchifyCaps), and a Cross-Agreement Routing (CAR) block. First, the MSRB extracts diverse multi-scale feature representations from input images, preserving both fine-grained details and global contextual information. Second, the PatchifyCaps partitions these multi-scale features into primary capsules using a uniform patch size, equipping the model with the ability to learn from diverse receptive fields. Finally, the CAR block adaptively routes the multi-scale capsules by identifying cross-scale prediction pairs with maximum agreement. Unlike the simple concatenation of multiple self-routing blocks, CAR ensures that only the most coherent capsules (best part-to-whole pairs) contribute to the final voting. Our proposed MSPCaps achieves remarkable scalability and superior robustness, consistently surpassing multiple baseline methods in terms of classification accuracy, with configurations ranging from a highly efficient Tiny model (344.3K parameters) to a powerful Large model (10.9M parameters), highlighting its potential in advancing feature representation learning.

Low-frame-rate (LFR) Multi-Object Tracking (MOT) is crucial for efficient tracking on edge devices, as it significantly reduces computational and storage demands. However, existing trackers struggle in LFR settings due to large temporal gaps, extreme appearance changes, and motion non-linearity. While Graph Neural Network (GNN)-based trackers are effective at associating objects across these gaps, most operate offline, which prevents their use for online tracking. To address these limitations, we propose GLoMOT, a novel online GNN-based Low-Frame-Rate Multi-Object Tracker designed for robust performance in LFR videos. To bridge the large temporal gaps, we introduce a Dynamic Node Buffer Pool. This acts as a long-term memory, caching the states of absent objects to enable their robust re-association. To tackle extreme motion uncertainty, we propose an adaptive context-aware module that dynamically adjusts the weights of positional and appearance features, generating more robust features for predicting node connections. Furthermore, we propose a pseudo-depth feature calculation method. This provides the GNN with critical geometric context, which helps resolve spatial ambiguity arising from occlusions. Extensive experiments on several public MOT benchmarks, including DanceTrack, MOT17, and VisDrone, demonstrate GLoMOT's effectiveness and superiority, particularly in challenging Low-Frame-Rate conditions.

Feed-forward 3D reconstruction from sparse, low-resolution (LR) images is a crucial capability for real-world applications, such as autonomous driving and embodied AI. However, existing methods often fail to recover fine texture details. This limitation stems from the inherent lack of high-frequency information in LR inputs. To address this, we propose SRSplat, a feed-forward framework that reconstructs high-resolution 3D scenes from only a few LR views. Our main insight is to compensate for the deficiency of texture information by jointly leveraging external high-quality reference images and internal texture cues. We first construct a scene-specific reference gallery, generated for each scene using Multimodal Large Language Models (MLLMs) and diffusion models. To integrate this external information, we introduce the Reference-Guided Feature Enhancement (RGFE) module, which aligns and fuses features from the LR input images and their reference twin image. Subsequently, we train a decoder to predict the Gaussian primitives using the multi-view fused feature obtained from RGFE. To further refine predicted Gaussian primitives, we introduce Texture-Aware Density Control (TADC), which adaptively adjusts Gaussian density based on the internal texture richness of the LR inputs. Extensive experiments demonstrate that our SRSplat outperforms existing methods on various datasets, including RealEstate10K, ACID, and DTU, and exhibits strong cross-dataset and cross-resolution generalization capabilities.

Vision Foundation Models (VFMs), while powerful, often struggle in Remote Sensing (RS) segmentation tasks when combined with existing Parameter-Efficient Fine-Tuning (PEFT) methods. We observe that this limitation primarily arises from their inability to effectively handle the pervasive artifacts in RS imagery. To address this, we introduce Earth-Adapter, the first PEFT method specifically designed for RS artifact mitigation. Earth-Adapter introduces a novel Frequency-Guided Mixture of Adapters (MoA) approach, structured around a ''divide and conquer" strategy. It first utilizes Discrete Fourier Transformation (DFT) to "divide" features into distinct frequency components, thereby effectively isolating artifact-related information from semantic signals. Subsequently, to ''conquer" these artifact, MoA independently optimizes features within different subspaces and dynamically assigns weights via a router to aggregate the refined representations. This enables adaptive refinement of the VFM’s representation space to mitigate the impact of artifacts. This simple yet highly effective PEFT method demonstrably mitigates artifacts and significantly enhances VFMs performance on RS segmentation tasks. Extensive experiments demonstrate Earth-Adapter's effectiveness on in-domain semantic segmentation (SS), as well as Domain Adaptive (DA) and Domain Generalized (DG) semantic segmentation tasks. Compared with the baseline Rein, Earth-Adapter significantly improves mIoU by 1.2% in SS, 9.0% in DA, and 3.1% in DG benchmarks. Our code and weights will be released soon.

Reconstructing complete and interactive 3D scenes remains a fundamental challenge in computer vision and robotics, particularly due to persistent object occlusions and limited sensor coverage. Even multi-view observations from a single scene scan often fail to capture the full structural details. Existing approaches typically rely on multi-stage pipelines—such as segmentation, background completion, and inpainting—or require per-object dense scanning, both of which are error-prone, and not easily scalable. We propose IGFuse, a novel framework that reconstructs interactive Gaussian scene by fusing observations from multiple scans, where natural object rearrangement between captures reveal previously occluded regions. Our method constructs segmentation-aware Gaussian fields and enforces bi-directional photometric and semantic consistency across scans. To handle spatial misalignments, we introduce a pseudo-intermediate scene state for symmetric alignment, alongside collaborative co-pruning strategies to refine geometry. IGFuse enables high-fidelity rendering and object-level scene manipulation without dense observations or complex pipelines. Extensive experiments validate the framework’s strong generalization to novel scene configurations, demonstrating its effectiveness for real-world 3D reconstruction and real-to-simulation transfer.

Recent advances in video generation have made it possible to produce visually compelling videos, with wide-ranging applications in content creation, entertainment, and virtual reality. However, most existing diffusion transformer based video generation models are limited to low-resolution outputs (

Multi-person pose estimation in real-world scenarios remains a challenging task due to frequent occlusions, scale variations, and complex human interactions. Existing methods often rely on fixed keypoint association patterns that fail to capture the dynamic and context-dependent nature of human body topologies, leading to misalignment and false detections. In this work, we propose a topology-aware dynamic association framework that adaptively models inter-keypoint relationships conditioned on local context and pose topology. The proposed framework comprises three stages: a human-to-keypoint detection module for coarse localization, a dynamic keypoint association module that learns flexible connectivity patterns between joints, and a fine-grained refinement module for precise pose adjustment. By integrating topological priors into dynamic learning and multi-stage optimization, our proposed method effectively mitigates the issues caused by occlusions and overlapping instances. Extensive experiments on benchmark datasets demonstrate that our approach achieves state-of-the-art performance, especially in crowded and occlusion-heavy scenes.

White-Light Imaging (WLI) is the standard for endoscopic cancer screening, but Narrow-Band Imaging (NBI) offers superior diagnostic details. A key challenge is transferring knowledge from NBI to enhance WLI-only models, yet existing methods are critically hampered by their reliance on paired NBI-WLI images of the same lesion, a costly and often impractical requirement that leaves vast amounts of clinical data untapped. In this paper, we break this paradigm by introducing PaGKD, a novel Pairing-free Group-level Knowledge Distillation framework that that enables effective cross-modal learning using unpaired WLI and NBI data. Instead of forcing alignment between individual, often semantically mismatched image instances, PaGKD operates at the group level to distill more complete and compatible knowledge across modalities. Central to PaGKD are two complementary modules: (1) Group-level Prototype Distillation (GKD-Pro) distills compact group representations by extracting modality-invariant semantic prototypes via shared lesion-aware queries; (2) Group-level Dense Distillation (GKD-Den) performs dense cross-modal alignment by guiding group-aware attention with activation-derived relation maps. Together, these modules enforce global semantic consistency and local structural coherence without requiring image-level correspondence. Extensive experiments on four clinical datasets demonstrate that PaGKD consistently and significantly outperforms state-of-the-art methods, boosting AUC by 3.3%, 1.1%, 2.8%, and 3.2%, respectively, establishing a new direction for cross-modal learning from unpaired data.

Leveraging intrinsic data priors is critical for effective data recovery. However, existing approaches often struggle to achieve theoretical guarantees, strong performance, and computational efficiency simultaneously. In this paper, we introduce a novel Representative Coefficient Correlated Total Variation (RCCTV) regularizer that captures the recently observed low-rank and local smoothness properties of the representative coefficient tensor derived from a low-rank decomposition. RCCTV regularizer offers three key advantages: (1) it operates on a compact representative coefficient image significantly smaller than the original data, enabling highly efficient optimization; (2) it jointly enforces low-rankness and spatial smoothness through a single regularizer, eliminating the need for trade-off parameters; and (3) when integrated into a robust PCA framework (i.e., RCCTV-RPCA model), it admits provable exact recovery under mild conditions. To solve the resulting model, we develop an efficient ADMM-based algorithm accelerated via fast Fourier transform. Extensive experiments on both synthetic and real-world datasets demonstrate that the RCCTV-RPCA model achieves state-of-the-art accuracy while running significantly faster. Our code and Supplementary Material are available at https://github.com/mendy-2013/RCCTV.

Existing RGB-T salient object detection methods predominantly rely on manually aligned and annotated datasets, struggling to handle real-world scenarios with raw, unaligned RGB-T image pairs. In practical applications, due to significant cross-modal disparities such as spatial misalignment, scale variations, and viewpoint shifts, the performance of current methods drastically deteriorates on unaligned datasets. To address this issue, we propose an efficient RGB-T SOD method for real-world unaligned image pairs, termed Thin-Plate Spline-driven Semantic Correlation Learning Network (TPS-SCL). We employ a dual-stream MobileViT as the encoder, combined with efficient Mamba scanning mechanisms, to effectively model correlations between the two modalities while maintaining low parameter counts and computational overhead. To suppress interference from redundant background information during alignment, we design a Semantic Correlation Constraint Module (SCCM) to hierarchically constrain salient features. Furthermore, we introduce a Thin-Plate Spline Alignment Module (TPSAM) to mitigate spatial discrepancies between modalities. Additionally, a Cross-Modal Correlation Module (CMCM) is incorporated to fully explore and integrate inter-modal dependencies, enhancing detection performance. Extensive experiments on various datasets demonstrate that TPS-SCL attains state-of-the-art (SOTA) performance among existing lightweight SOD methods and outperforms mainstream RGB-T SOD approaches.

Forecasting 3D human motion is an important embodiment of fine-grained understanding and cognition of human behavior by artificial agents. Current approaches excessively rely on implicit network modeling of spatiotemporal relationships and motion characteristics, falling into the passive learning trap that results in redundant and monotonous 3D coordinate information acquisition while lacking actively guided explicit learning mechanisms. To overcome these issues, we propose an Active Perceptual Strategy (APS) for human motion prediction, leveraging quotient space representations to explicitly encode motion properties while introducing auxiliary learning objectives to strengthen spatio-temporal modeling. Specifically, we first design a data perception module that projects poses into the quotient space, decoupling motion geometry from coordinate redundancy. By jointly encoding tangent vectors and Grassmann projections, this module simultaneously achieves geometric dimension reduction, semantic decoupling, and dynamic constraint enforcement for effective motion pose characterization. Furthermore, we introduce a network perception module that actively learns spatio-temporal dependencies through restorative learning. This module deliberately masks specific joints or injects noise to construct auxiliary supervision signals. A dedicated auxiliary learning network is designed to actively adapt and learn from perturbed information. Notably, APS is model agnostic and can be integrated with different prediction models to enhance active perceptual.The experimental results demonstrate that our method achieves the new state-of-the-art, outperforming existing methods by large margins: 16.3% on H3.6M, 13.9% on CMU Mocap, and 10.1% on 3DPW.

In-context learning (ICL) offers a promising paradigm for universal medical image analysis, enabling models to perform diverse image processing tasks without retraining. However, current ICL models for medical imaging remain limited in two critical aspects: they cannot simultaneously achieve high-fidelity predictions and global anatomical understanding, and there is no unified model trained across diverse medical imaging tasks (e.g., segmentation and enhancement) and anatomical regions. As a result, the full potential of ICL in medical imaging remains underexplored. Thus, we present Medverse, a universal ICL model for 3D medical imaging, trained on 22 datasets covering diverse tasks in universal image segmentation, transformation, and enhancement across multiple organs, imaging modalities, and clinical centers. Medverse employs a next-scale autoregressive in-context learning framework that progressively refines predictions from coarse to fine, generating consistent, full-resolution volumetric outputs and enabling multi-scale anatomical awareness. We further propose a blockwise cross-attention module that facilitates long-range interactions between context and target inputs while preserving computational efficiency through spatial sparsity. Medverse is extensively evaluated on a broad collection of held-out datasets covering previously unseen clinical centers, organs, species, and imaging modalities. Results demonstrate that Medverse substantially outperforms existing ICL baselines and establishes a novel paradigm for in-context learning.

Reflection removal of a single image remains a highly challenging task due to the complex entanglement between target scenes and unwanted reflections. Despite significant progress, existing methods are hindered by the scarcity of high-quality, diverse data and insufficient restoration priors, resulting in limited generalization across various real-world scenarios. In this paper, we propose Dereflection Any Image, a comprehensive solution with an efficient data preparation pipeline and a generalizable model for robust reflection removal. First, we introduce a dataset named Diverse Reflection Removal (DRR) created by randomly rotating reflective mediums in target scenes, enabling variation of reflection angles and intensities, and setting a new benchmark in scale, quality, and diversity. Second, we propose a diffusion-based framework with one-step diffusion for deterministic outputs and fast inference. To ensure stable learning, we design a three-stage progressive training strategy, including reflection-invariant finetuning to encourage consistent outputs across varying reflection patterns that characterize our dataset. Extensive experiments show that our method achieves SOTA performance on both common benchmarks and challenging in-the-wild images, showing superior generalization across diverse real-world scenes.

Low-light object detection faces significant challenges due to the substantial domain shift between normal-light and low-light conditions. Prior works often enhance low-light images before detection, but this preprocessing can introduce artifacts that degrade detection performance since it focuses on human visual quality rather than task-specific features. Other methods incorporate illumination-aware modules for low-light feature learning, yet their scalability is limited by the scarcity of annotated low-light datasets. To overcome these limitations, we propose a unified Dual-Level Domain Adaptation (DLDA) framework that jointly addresses image-level and feature-level domain discrepancies. Specifically, we introduce a luminance-aware contrastive translation module that synthesizes target-style low-light images while preserving structural details, enabling effective image-level adaptation. Building on this, we further design a multi-scale conditional adversarial alignment strategy that promotes semantic consistency across feature hierarchies to enhance domain-invariant feature extraction. Extensive experiments on multiple low-light detection benchmarks demonstrate that DLDA achieves state-of-the-art performance, exhibiting strong robustness and generalization.

While 3D Gaussian Splatting (3DGS) excels at real-time rendering of standard scenes, it struggles to reconstruct underwater environments due to severe challenges such as light scattering, color attenuation, and sparse coverage of Gaussian kernels in far-field aqueous regions. To address this, we introduce AquaSplatting, a hybrid framework that combines explicit and implicit modeling methods for robust underwater scene reconstruction. Our dual-branch architecture employs 3DGS in a geometry-guided branch to model solid surfaces like the seabed, while a medium-aware branch uses a compact, view-dependent MLP to represent volumetric water effects. Furthermore, a neural underwater hybrid rendering mechanism adaptively fuses these two representations based on accumulated opacity. Thanks to this dual-branch framework, our method can also synthesize restored images without water medium. To enhance efficiency, our proposed engagement-based pruning (EBP) strategy quantifies each Gaussian's contribution by accumulating its image-space gradients over multiple frames, enabling the principled removal of primitives with negligible impact. The entire framework is optimized using a comprehensive loss function that integrates photometric, exposure, semantic, and depth priors to maximize visual fidelity. Experiments on challenging underwater datasets demonstrate that AquaSplatting achieves the state-of-the-art in reconstruction quality surpassing prior methods while maintaining real-time performance.

Reconstructing dynamic scenes has long been a challenging task in 3D vision. Previous mainstream methods based on 3D Gaussian Splatting typically employ a single deformation field to directly model spatiotemporal changes. However, such one-step deformation struggles to capture diverse and complex motion patterns. To address this limitation, we propose decomposing the one-step deformation into a multi-step process, where each step is represented by a deformation layer. Additionally, we introduce a weight prediction mechanism for each layer to control the extent of deformation at every step. We provide two types of deformation layers based on implicit and explicit approaches. Moreover, while the deformation layer is time-conditioned, the Gaussians' behavior may still be influenced by their time-invariant properties. Therefore, we propose a fully time-agnostic scale modulation block to modulate the scaling changes of Gaussians. Extensive experiments on D-NeRF, Neu3D, and HyperNeRF demonstrate that our method achieves state-of-the-art performance.

This work focuses on multi-shot semi-supervised video object segmentation (MVOS), which aims at segmenting the target object indicated by an initial mask throughout a video with multiple shots. While existing VOS methods mainly focus on single-shot videos, they often fail to handle shot discontinuities, thereby limiting their real-world applicability. Furthermore, the lack of annotated multi-shot data poses a major challenge for MVOS research. To address these issues, we propose a transition mimicking data augmentation strategy (TMA) that enables cross-shot generalization using single-shot data, and a transition-aware method, Segment Anything Across Shots (SAAS), which detects and comprehends shot transitions during inference. To support evaluation and future study in MVOS, we introduce Cut-VOS, a new MVOS benchmark with dense mask annotations, diverse object categories, and high-frequency transitions. Extensive experiments on YouMVOS and Cut-VOS demonstrate that the proposed SAAS achieves state-of-the-art performance by effectively mimicking, understanding, and segmenting across complex transitions.