NeurIPS 2025 accepted papers on 3D vision
265 of the 5,812 papers accepted at NeurIPS 2025 (4.6%) are on 3D vision, compared with 4.8% at NeurIPS 2024. They fall into 6 topics.
See NeurIPS 2025 in the interactive map
Working on something in this area? Describe your idea in scime atlas to see the closest accepted papers by meaning and how crowded the spot has become.
Topics and papers
3D generation · 93 papers
- Learning from Videos for 3D World: Enhancing MLLMs with 3D Vision Geometry Priors (14 citations)
- From Flatland to Space: Teaching Vision-Language Models to Perceive and Reason in 3D (12 citations)
- EgoExOR: An Ego-Exo-Centric Operating Room Dataset for Surgical Activity Understanding (6 citations)
- PartCrafter: Structured 3D Mesh Generation via Compositional Latent Diffusion Transformers (6 citations)
- SceneWeaver: All-in-One 3D Scene Synthesis with an Extensible and Self-Reflective Agent (5 citations)
- X-Scene: Large-Scale Driving Scene Generation with High Fidelity and Flexible Controllability (5 citations)
- InternScenes: A Large-scale Simulatable Indoor Scene Dataset with Realistic Layouts (4 citations)
- MindJourney: Test-Time Scaling with World Models for Spatial Reasoning (4 citations)
- PhysX-3D: Physical-Grounded 3D Asset Generation (4 citations)
- Reconstruct, Inpaint, Test-Time Finetune: Dynamic Novel-view Synthesis from Monocular Videos (4 citations)
- 4D-LRM: Large Space-Time Reconstruction Model From and To Any View at Any Time (3 citations)
- Direct Numerical Layout Generation for 3D Indoor Scene Synthesis via Spatial Reasoning (3 citations)
and 81 more in the interactive map.
Gaussian splatting · 75 papers
- Feed-Forward Bullet-Time Reconstruction of Dynamic Scenes from Monocular Videos (6 citations)
- FlexWorld: Progressively Expanding 3D Scenes for Flexible-View Exploration (6 citations)
- Optimized Minimal 3D Gaussian Splatting (5 citations)
- Gaussian Herding across Pens: An Optimal Transport Perspective on Global Gaussian Reduction for 3DGS (4 citations)
- LangSplatV2: High-dimensional 3D Language Gaussian Splatting with 450+ FPS (4 citations)
- ZPressor: Bottleneck-Aware Compression for Scalable Feed-Forward 3DGS (4 citations)
- DGS-LRM: Real-Time Deformable 3D Gaussian Reconstruction From Monocular Videos (3 citations)
- Flux4D: Flow-based Unsupervised 4D Reconstruction (3 citations)
- HyRF: Hybrid Radiance Fields for Memory-efficient and High-quality Novel View Synthesis (3 citations)
- Instant4D: 4D Gaussian Splatting in Minutes (3 citations)
- NerfBaselines: Consistent and Reproducible Evaluation of Novel View Synthesis Methods (3 citations)
- AtlasGS: Atlanta-world Guided Surface Reconstruction with Implicit Structured Gaussians (2 citations)
and 63 more in the interactive map.
Depth and stereo · 39 papers
- MoGe-2: Accurate Monocular Geometry with Metric Scale and Sharp Details (27 citations)
- Cameras as Relative Positional Encoding (10 citations)
- VGGT-SLAM: Dense RGB SLAM Optimized on the SL(4) Manifold (8 citations)
- TAPIP3D: Tracking Any Point in Persistent 3D Geometry (6 citations)
- UFM: A Simple Path towards Unified Dense Correspondence with Flow (6 citations)
- DualFocus: Depth from Focus with Spatio-Focal Dual Variational Constraints (2 citations)
- Event-Driven Dynamic Scene Depth Completion (2 citations)
- UltraLED: Learning to See Everything in Ultra-High Dynamic Range Scenes (2 citations)
- Are Pixel-Wise Metrics Reliable for Computerized Tomography Reconstruction? (1 citation)
- EAG3R: Event-Augmented 3D Geometry Estimation for Dynamic and Extreme-Lighting Scenes (1 citation)
- Evaluating Robustness of Monocular Depth Estimation with Procedural Scene Perturbations (1 citation)
- GeoSVR: Taming Sparse Voxels for Geometrically Accurate Surface Reconstruction (1 citation)
and 27 more in the interactive map.
Face and shape · 30 papers
- PolyPose: Deformable 2D/3D Registration via Polyrigid Transformations (3 citations)
- VASA-3D: Lifelike Audio-Driven Gaussian Head Avatars from a Single Image (2 citations)
- BecomingLit: Relightable Gaussian Avatars with Hybrid Neural Shading (1 citation)
- FLOWING: Implicit Neural Flows for Structure-Preserving Morphing (1 citation)
- GeneMAN: Generalizable Single-Image 3D Human Reconstruction from Multi-Source Human Data (1 citation)
- Grids Often Outperform Implicit Neural Representation at Compressing Dense Signals (1 citation)
- RaySt3R: Predicting Novel Depth Maps for Zero-Shot Object Completion (1 citation)
- Template-Guided 3D Molecular Pose Generation via Flow Matching and Differentiable Optimization (1 citation)
- Wukong's 72 Transformations: High-fidelity Textured 3D Morphing via Flow Models (1 citation)
- 3DID: Direct 3D Inverse Design for Aerodynamics with Physics-Aware Optimization
- AMBER: Adaptive Mesh Generation by Iterative Mesh Resolution Prediction
- ARMesh: Autoregressive Mesh Generation via Next-Level-of-Detail Prediction
and 18 more in the interactive map.
Human pose estimation · 19 papers
- BEDLAM2.0: Synthetic humans and cameras in motion (7 citations)
- Coarse-to-Fine 3D Part Assembly via Semantic Super-Parts and Symmetry-Aware Pose Estimation (1 citation)
- Orient Anything V2: Unifying Orientation and Rotation Understanding (1 citation)
- PyraMotion: Attentional Pyramid-Structured Motion Integration for Co-Speech 3D Gesture Synthesis (1 citation)
- Whole-Body Conditioned Egocentric Video Prediction (1 citation)
- 3D Equivariant Visuomotor Policy Learning via Spherical Projection
- 3D Human Pose Estimation with Muscles
- Cycle-Sync: Robust Global Camera Pose Estimation through Enhanced Cycle-Consistent Synchronization
- HO-Cap: A Capture System and Dataset for 3D Reconstruction and Pose Tracking of Hand-Object Interaction
- NoPo-Avatar: Generalizable and Animatable Avatars from Sparse Inputs without Human Poses
- PandaPose: 3D Human Pose Lifting from a Single Image via Propagating 2D Pose Prior to 3D Anchor Space
- RAPTR: Radar-based 3D Pose Estimation using Transformer
and 7 more in the interactive map.
NeRF and radiance fields · 9 papers
- LinPrim: Linear Primitives for Differentiable Volumetric Rendering (2 citations)
- I2-NeRF: Learning Neural Radiance Fields Under Physically-Grounded Media Interactions (1 citation)
- Learning Neural Exposure Fields for View Synthesis (1 citation)
- ROGR: Relightable 3D Objects using Generative Relighting (1 citation)
- Abstract Rendering: Certified Rendering Under 3D Semantic Uncertainty
- Can NeRFs "See" without Cameras?
- KaRF: Weakly-Supervised Kolmogorov-Arnold Networks-based Radiance Fields for Local Color Editing
- OpenHype: Hyperbolic Embeddings for Hierarchical Open-Vocabulary Radiance Fields
- Optimize the Unseen - Fast NeRF Cleanup with Free Space Prior