ICCV 2025 accepted papers on 3D vision
562 of the 2,701 papers accepted at ICCV 2025 (20.8%) are on 3D vision, compared with 20.9% at ICCV 2023. They fall into 6 topics.
See ICCV 2025 in the interactive map
Submitting to ICCV? Checklist →
Working on something in this area? Describe your idea in scime atlas to see the closest papers by meaning and how crowded the spot has become.
Topics and papers
3D generation · 161 papers
- Stable Virtual Camera: Generative View Synthesis with Diffusion Models (138 citations)
- MV-Adapter: Multi-View Consistent Image Generation Made Easy (102 citations)
- PartField: Learning 3D Feature Fields for Part Segmentation and Beyond (61 citations)
- PhysTwin: Physics-Informed Reconstruction and Simulation of Deformable Objects from Videos (51 citations)
- Ross3D: Reconstructive Visual Instruction Tuning with 3D-Awareness (49 citations)
- HERMES: A Unified Self-Driving World Model for Simultaneous 3D Scene Understanding and Generation (45 citations)
- Bolt3D: Generating 3D Scenes in Seconds (41 citations)
- DiST-4D: Disentangled Spatiotemporal Diffusion with Metric Depth for 4D Driving Scene Generation (37 citations)
- VistaDream: Sampling multiview consistent images for single-view scene reconstruction (36 citations)
- REPARO: Compositional 3D Assets Generation with Differentiable 3D Layout Alignment (35 citations)
- PhysSplat: Efficient Physics Simulation for 3D Scenes via MLLM-Guided Gaussian Splatting (33 citations)
- WonderTurbo: Generating Interactive 3D World in 0.72 Seconds (31 citations)
and 149 more in the interactive map.
Gaussian splatting · 149 papers
- Shape of Motion: 4D Reconstruction from a Single Video (186 citations)
- EvaGaussians: Event Stream Assisted Gaussian Splatting from Blurry Images (63 citations)
- Long-LRM: Long-sequence Large Reconstruction Model for Wide-coverage Gaussian Splats (63 citations)
- EVER: Exact Volumetric Ellipsoid Rendering for Real-time View Synthesis (46 citations)
- Radiant Foam: Real-Time Differentiable Ray Tracing (41 citations)
- GaussianOcc: Fully Self-supervised and Efficient 3D Occupancy Estimation with Gaussian Splatting (39 citations)
- MEGA: Memory-Efficient 4D Gaussian Splatting for Dynamic Scenes (39 citations)
- 3DGS-LM: Faster Gaussian-Splatting Optimization with Levenberg-Marquardt (38 citations)
- Compression of 3D Gaussian Splatting with Optimized Feature Planes and Standard Video Codecs (37 citations)
- ReconDreamer++: Harmonizing Generative and Reconstructive Models for Driving Scene Representation (37 citations)
- FreeSplatter: Pose-free Gaussian Splatting for Sparse-view 3D Reconstruction (33 citations)
- BillBoard Splatting (BBSplat): Learnable Textured Primitives for Novel View Synthesis (31 citations)
and 137 more in the interactive map.
Face and shape · 89 papers
- MeshAnything V2: Artist-Created Mesh Generation with Adjacent Mesh Tokenization (95 citations)
- Hi3DGen: High-fidelity 3D Geometry Generation from Images via Normal Bridging (93 citations)
- SparseFlex: High-Resolution and Arbitrary-Topology 3D Shape Modeling (74 citations)
- Amodal3R: Amodal 3D Reconstruction from Occluded 2D Images (48 citations)
- 3DRealCar: An In-the-wild RGB-D Car Dataset with 360-degree Views (43 citations)
- Avat3r: Large Animatable Gaussian Reconstruction Model for High-fidelity 3D Head Avatars (40 citations)
- LHM: Large Animatable Human Reconstruction Model for Single Image to 3D in Seconds (34 citations)
- GUAVA: Generalizable Upper Body 3D Gaussian Avatar (20 citations)
- Nautilus: Locality-aware Autoencoder for Scalable Mesh Generation (18 citations)
- 3D Mesh Editing using Masked LRMs (16 citations)
- DreamDance: Animating Human Images by Enriching 3D Geometry Cues from 2D Poses (15 citations)
- MeshLLM: Empowering Large Language Models to Progressively Understand and Generate 3D Mesh (13 citations)
and 77 more in the interactive map.
Depth and stereo · 86 papers
- Easi3R: Estimating Disentangled Motion from DUSt3R Without Training (76 citations)
- Marigold-DC: Zero-Shot Monocular Depth Completion with Guided Diffusion (56 citations)
- St4RTrack: Simultaneous 4D Reconstruction and Tracking in the World (56 citations)
- RayZer: A Self-supervised Large View Synthesis Model (52 citations)
- Geo4D: Leveraging Video Generators for Geometric 4D Scene Reconstruction (50 citations)
- Dynamic Point Maps: A Versatile Representation for Dynamic 3D Reconstruction (30 citations)
- POMATO: Marrying Pointmap Matching with Temporal Motions for Dynamic 3D Reconstruction (23 citations)
- OMNI-DC: Highly Robust Depth Completion with Multiresolution Depth Integration (17 citations)
- BANet: Bilateral Aggregation Network for Mobile Stereo Matching (16 citations)
- AnyCalib: On-Manifold Learning for Model-Agnostic Single-View Camera Calibration (15 citations)
- S2M2: Scalable Stereo Matching Model for Reliable Depth Estimation (15 citations)
- BridgeDepth: Bridging Monocular and Stereo Reasoning with Latent Alignment (14 citations)
and 74 more in the interactive map.
Human pose estimation · 65 papers
- DPoser-X: Diffusion Model as Robust 3D Whole-body Human Pose Prior (14 citations)
- egoPPG: Heart Rate Estimation from Eye-Tracking Cameras in Egocentric Systems to Benefit Downstream Vision Tasks (14 citations)
- Generative Zoo (12 citations)
- A Structure-aware and Motion-adaptive Framework for 3D Human Pose Estimation with Mamba (11 citations)
- Bring Your Rear Cameras for Egocentric 3D Human Pose Estimation (10 citations)
- CleanPose: Category-Level Object Pose Estimation via Causal Learning and Knowledge Distillation (10 citations)
- RePoseD: Efficient Relative Pose Estimation With Known Depth Information (9 citations)
- PersPose: 3D Human Pose Estimation with Perspective Encoding and Perspective Rotation (7 citations)
- PoseAnchor: Robust Root Position Estimation for 3D Human Pose Estimation (7 citations)
- 6DOPE-GS: Online 6D Object Pose Estimation using Gaussian Splatting (6 citations)
- CapeLLM: Support-Free Category-Agnostic Pose Estimation with Multimodal Large Language Models (6 citations)
- High-Resolution Spatiotemporal Modeling with Global-Local State Space Models for Video-Based Human Pose Estimation (6 citations)
and 53 more in the interactive map.
NeRF and radiance fields · 12 papers
- Bokehlicious: Photorealistic Bokeh Rendering with Controllable Apertures (8 citations)
- UnMix-NeRF: Spectral Unmixing Meets Neural Radiance Fields (3 citations)
- AdaptiveAE: An Adaptive Exposure Strategy for HDR Capturing in Dynamic Scenes (2 citations)
- E-NeMF: Event-based Neural Motion Field for Novel Space-time View Synthesis of Dynamic Scenes (2 citations)
- Relative Illumination Fields: Learning Medium and Light Independent Underwater Scenes (2 citations)
- Robust Low-light Scene Restoration via Illumination Transition (2 citations)
- Inverse Image-Based Rendering for Light Field Generation from Single Images (1 citation)
- Learning Neural Scene Representation from iToF Imaging (1 citation)
- Learning Null Geodesics for Gravitational Lensing Rendering in General Relativity (1 citation)
- A View-consistent Sampling Method for Regularized Training of Neural Radiance Fields
- Exploring View Consistency for Scene-Adaptive Low-Light Light Field Image Enhancement
- LightCity: An Urban Dataset for Outdoor Inverse Rendering and Reconstruction under Multi-illumination Conditions
