ICCV 2025
自动驾驶 · 具身智能论文归档

Where, What, Why: Towards Explainable Driver Attention Prediction

Interaction-Merged Motion Planning: Effectively Leveraging Diverse Motion Datasets for Robust Planning

SRefiner: Soft-Braid Attention for Multi-Agent Trajectory Refinement

Distilling Diffusion Models to Efficient 3D LiDAR Scene Completion

AutoOcc: Automatic Open-Ended Semantic Occupancy Annotation via Vision-Language Guided Gaussian Splatting

Multi-View 3D Point Tracking

Diving into the Fusion of Monocular Priors for Generalized Stereo Matching

Counting Stacked Objects

CorrCLIP: Reconstructing Patch Correlations in CLIP for Open-Vocabulary Semantic Segmentation

E-SAM: Training-Free Segment Every Entity Model

Easy3D: A Simple Yet Effective Method for 3D Interactive Segmentation

Towards Foundational Models for Single-Chip Radar

ForestFormer3D: A Unified Framework for End-to-End Segmentation of Forest LiDAR 3D Point Clouds

ReTracker: Exploring Image Matching for Robust Online Any Point Tracking

Test-Time Prompt Tuning for Zero-Shot Depth Completion

DiffRefine: Diffusion-based Proposal Specific Point Cloud Densification for Cross-Domain Object Detection

Boosting Domain Generalized and Adaptive Detection with Diffusion Models: Fitness, Generalization, and Transferability

Multispectral Demosaicing via Dual Cameras

CoTracker3: Simpler and Better Point Tracking by Pseudo-Labelling Real Videos

StableDepth: Scene-Consistent and Scale-Invariant Monocular Depth

Find Any Part in 3D

Unsupervised Joint Learning of Optical Flow and Intensity with Event Cameras

ETCH: Generalizing Body Fitting to Clothed Humans via Equivariant Tightness

FlashDepth: Real-time Streaming Video Depth Estimation at 2K Resolution

Few-Shot Pattern Detection via Template Matching and Regression

Interpretable point cloud classification using multiple instance learning

Prior2Former - Evidential Modeling of Mask Transformers for Assumption-Free Open-World Panoptic Segmentation

OmniSAM: Omnidirectional Segment Anything Model for UDA in Panoramic Semantic Segmentation

CoopTrack: Exploring End-to-End Learning for Efficient Cooperative Sequential Perception

BridgeDepth: Bridging Monocular and Stereo Reasoning with Latent Alignment

Lidar Waveforms are Worth 40x128x33 Words

Adaptive Dual Uncertainty Optimization: Boosting Monocular 3D Object Detection under Test-Time Shifts

CoST: Efficient Collaborative Perception From Unified Spatiotemporal Perspective

DAP-MAE: Domain-Adaptive Point Cloud Masked Autoencoder for Effective Cross-Domain Learning

UniDxMD: Towards Unified Representation for Cross-Modal Unsupervised Domain Adaptation in 3D Semantic Segmentation

Height-Fidelity Dense Global Fusion for Multi-modal 3D Object Detection

SMSTracker: Tri-path Score Mask Sigma Fusion for Multi-Modal Tracking

Stereo Any Video: Temporally Consistent Stereo Matching

When Confidence Fails: Revisiting Pseudo-Label Selection in Semi-supervised Semantic Segmentation

Learning Robust Stereo Matching in the Wild with Selective Mixture-of-Experts

Structure Matters: Revisiting Boundary Refinement in Video Object Segmentation

CounterPC: Counterfactual Feature Realignment for Unsupervised Domain Adaptation on Point Clouds

ATCTrack: Aligning Target-Context Cues with Dynamic Target States for Robust Vision-Language Tracking

PriOr-Flow: Enhancing Primitive Panoramic Optical Flow with Orthogonal View

GameFactory: Creating New Games with Generative Interactive Videos

From Image to Video: An Empirical Study of Diffusion Representations

VMem: Consistent Interactive Video Scene Generation with Surfel-Indexed View Memory

Moto: Latent Motion Token as the Bridging Language for Learning Robot Manipulation from Videos

Combinative Matching for Geometric Shape Assembly

DexVLG: Dexterous Vision-Language-Grasp Model at Scale

Contact-Aware Amodal Completion for Human-Object Interaction via Multi-Regional Inpainting

MaskControl: Spatio-Temporal Control for Masked Motion Synthesis

DPoser-X: Diffusion Model as Robust 3D Whole-body Human Pose Prior

VolumetricSMPL: A Neural Volumetric Body Model for Efficient Interactions, Contacts, and Collisions

SceneMI: Motion In-betweening for Modeling Human-Scene Interaction

Towards Immersive Human-X Interaction: A Real-Time Framework for Physically Plausible Motion Synthesis

GENMO: A GENeralist Model for Human MOtion

UniPhys: Unified Planner and Controller with Diffusion for Flexible Physics-Based Character Control

DisCoRD: Discrete Tokens to Continuous Motion via Rectified Flow Decoding

A Plug-and-Play Physical Motion Restoration Approach for In-the-Wild High-Difficulty Motions

Online Reasoning Video Segmentation with Just-in-Time Digital Twins

Less is More: Empowering GUI Agent with Context-Aware Simplification

Embodied VideoAgent: Persistent Memory from Egocentric Videos and Embodied Sensors Enables Dynamic Scene Understanding

Visual Test-time Scaling for GUI Agent Grounding

Move to Understand a 3D Scene: Bridging Visual Grounding and Exploration for Efficient and Versatile Embodied Navigation

Embodied Image Captioning: Self-supervised Learning Agents for Spatially Coherent Image Descriptions

GT-Loc: Unifying When and Where in Images Through a Joint Embedding Space

Uncalibrated Structure from Motion on a Sphere

Image as an IMU: Estimating Camera Motion from a Single Motion-Blurred Image

Back on Track: Bundle Adjustment for Dynamic Scene Reconstruction

Forecasting Continuous Non-Conservative Dynamical Systems in SO(3)

Certifiably Optimal Anisotropic Rotation Averaging

Deterministic Object Pose Confidence Region Estimation

RePoseD: Efficient Relative Pose Estimation With Known Depth Information

RALoc: Enhancing Outdoor LiDAR Localization via Rotation Awareness

Registration beyond Points: General Affine Subspace Alignment via Geodesic Distance on Grassmann Manifold

BUFFER-X: Towards Zero-Shot Point Cloud Registration in Diverse Scenes

A Linear N-Point Solver for Structure and Motion from Asynchronous Tracks

Fast Globally Optimal and Geometrically Consistent 3D Shape Matching

SAC-GNC: SAmple Consensus for adaptive Graduated Non-Convexity

HccePose(BF): Predicting Front & Back Surfaces to Construct Ultra-Dense 2D-3D Correspondences for Pose Estimation

CoMatch: Dynamic Covisibility-Aware Transformer for Bilateral Subpixel-Level Semi-Dense Image Matching

On the Recovery of Cameras from Fundamental Matrices

EDM: Efficient Deep Feature Matching

SGAD: Semantic and Geometric-aware Descriptor for Local Feature Matching

Underwater Visual SLAM with Depth Uncertainty and Medium Modeling

CasP: Improving Semi-Dense Feature Matching Pipeline Leveraging Cascaded Correspondence Priors for Guidance

PLMP - Point-Line Minimal Problems for Projective SfM

A Constrained Optimization Approach for Gaussian Splatting from Coarsely-posed Images and Noisy Lidar Point Clouds

Kaleidoscopic Background Attack: Disrupting Pose Estimation with Multi-Fold Radial Symmetry Textures

SuperEvent: Cross-Modal Learning of Event-based Keypoint Detection for SLAM

Heavy Labels Out! Dataset Distillation with Label Space Lightening

ReCamMaster: Camera-Controlled Generative Rendering from A Single Video

Stable-Sim2Real: Exploring Simulation of Real-Captured 3D Data with Two-Stage Depth Diffusion

Guiding Diffusion-Based Articulated Object Generation by Partial Point Cloud Alignment and Physical Plausibility Constraints

Is Tracking Really More Challenging in First Person Egocentric Vision?

Learning Large Motion Estimation from Intermediate Representations with a High-Resolution Optical Flow Dataset Featuring Long-Range Dynamic Motion

AdvDreamer Unveils: Are Vision-Language Models Truly Ready for Real-World 3D Variations?

WonderPlay: Dynamic 3D Scene Generation from a Single Image and Actions

Go to Zero: Towards Zero-shot Motion Generation with Million-scale Data

RESCUE: Crowd Evacuation Simulation via Controlling SDM-United Characters

Benchmarking Egocentric Visual-Inertial SLAM at City Scale

MaterialMVP: Illumination-Invariant Material Generation via Multi-view PBR Diffusion

PolarAnything: Diffusion-based Polarimetric Image Synthesis

DSO: Aligning 3D Generators with Simulation Feedback for Physical Soundness

UnrealZoo: Enriching Photo-realistic Virtual Worlds for Embodied AI

Towards Scalable Spatial Intelligence via 2D-to-3D Data Lifting

TrajectoryCrafter: Redirecting Camera Trajectory for Monocular Videos via Diffusion Models

RayZer: A Self-supervised Large View Synthesis Model

EVER: Exact Volumetric Ellipsoid Rendering for Real-time View Synthesis

Self-Ensembling Gaussian Splatting for Few-Shot Novel View Synthesis

SceneSplat: Gaussian Splatting-based Scene Understanding with Vision-Language Pretraining

WIR3D: Visually-Informed and Geometry-Aware 3D Shape Abstraction

SparseFlex: High-Resolution and Arbitrary-Topology 3D Shape Modeling

SuperDec: 3D Scene Decomposition with Superquadrics Primitives

Radiant Foam: Real-Time Differentiable Ray Tracing

Long-LRM: Long-sequence Large Reconstruction Model for Wide-coverage Gaussian Splats

Human-in-the-Loop Local Corrections of 3D Scene Layouts via Infilling

Dynamic Point Maps: A Versatile Representation for Dynamic 3D Reconstruction

CObL: Toward Zero-Shot Ordinal Layering without User Prompting

MaGS: Reconstructing and Simulating Dynamic 3D Objects with Mesh-adsorbed Gaussian Splatting

Diorama: Unleashing Zero-shot Single-view 3D Indoor Scene Modeling

CHROME: Clothed Human Reconstruction with Occlusion-Resilience and Multiview-Consistency from a Single Image

MeshLLM: Empowering Large Language Models to Progressively Understand and Generate 3D Mesh

Blended Point Cloud Diffusion for Localized Text-guided Shape Editing

Geo4D: Leveraging Video Generators for Geometric 4D Scene Reconstruction

PRM: Photometric Stereo based Large Reconstruction Model

RayletDF: Raylet Distance Fields for Generalizable 3D Surface Reconstruction from Point Clouds or Gaussians

HiNeuS: High-fidelity Neural Surface Mitigating Low-texture and Reflective Ambiguity

Discontinuity-aware Normal Integration for Generic Central Camera Models

Explaining Human Preferences via Metrics for Structured 3D Reconstruction

Thermal Polarimetric Multi-view Stereo

AAA-Gaussians: Anti-Aliased and Artifact-Free 3D Gaussian Rendering

SAFT: Shape and Appearance of Fabrics from Template via Differentiable Physical Simulations from Monocular Video

FlowR: Flowing from Sparse to Dense 3D Reconstructions

Stochastic Gradient Estimation for Higher-Order Differentiable Rendering

HouseCrafter: Lifting Floorplans to 3D Scenes with 2D Diffusion Models

No Pose at All: Self-Supervised Pose-Free 3D Gaussian Splatting from Sparse Views

Sparfels: Fast Reconstruction from Sparse Unposed Imagery

Shape of Motion: 4D Reconstruction from a Single Video

DIMO: Diverse 3D Motion Generation for Arbitrary Objects

MEGA: Memory-Efficient 4D Gaussian Splatting for Dynamic Scenes

Self-Calibrating Gaussian Splatting for Large Field-of-View Reconstruction

Geometry Distributions

MoGA: 3D Generative Avatar Prior for Monocular Gaussian Avatar Reconstruction

Scaling Laws for Native Multimodal Models

Removing Cost Volumes from Optical Flow Estimators

Token Activation Map to Visually Explain Multimodal LLMs

Variance-Based Pruning for Accelerating and Compressing Trained Networks

Importance-Based Token Merging for Efficient Image and Video Generation

Knowledge Distillation for Learned Image Compression

Learning Streaming Video Representation via Multitask Training

LoftUp: Learning a Coordinate-Based Feature Upsampler for Vision Foundation Models

VRM: Knowledge Distillation via Virtual Relation Matching

One Trajectory, One Token: Grounded Video Tokenization via Panoptic Sub-object Trajectory

Scaling Language-Free Visual Representation Learning

Confound from All Sides, Distill with Resilience: Multi-Objective Adversarial Paths to Zero-Shot Robustness

What to Distill? Fast Knowledge Distillation with Adaptive Sampling

Corvid: Improving Multimodal Large Language Models Towards Chain-of-Thought Reasoning

MEMFOF: High-Resolution Training for Memory-Efficient Multi-Frame Optical Flow Estimation

Outlier-Aware Post-Training Quantization for Image Super-Resolution

Adversarial Distribution Matching for Diffusion Distillation Towards Efficient Image and Video Synthesis

EVEv2: Improved Baselines for Encoder-Free Vision-Language Models

WINS: Winograd Structured Pruning for Fast Winograd Convolution

Cross-Architecture Distillation Made Simple with Redundancy Suppression

Token-Efficient VLM: High-Resolution Image Understanding via Dynamic Region Proposal

Emulating Self-attention with Convolution for Efficient Image Super-Resolution

Beyond Losses Reweighting: Empowering Multi-Task Learning via the Generalization Perspective

Rectifying Magnitude Neglect in Linear Attention

Region-based Cluster Discrimination for Visual Representation Learning

The Scalability of Simplicity: Empirical Analysis of Vision-Language Learning with a Single Transformer

ResidualViT for Efficient Temporally Dense Video Encoding