ICLR 2026 accepted papers on 3D vision
208 of the 5,358 papers accepted at ICLR 2026 (3.9%) are on 3D vision, compared with 3.6% at ICLR 2025. They fall into 6 topics.
See ICLR 2026 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
Gaussian splatting · 66 papers
- Learning Physics-Grounded 4D Dynamics with Neural Gaussian Force Fields (1 citation)
- Sharp Monocular View Synthesis in Less Than a Second (1 citation)
- SSD-GS: Scattering and Shadow Decomposition for Relightable 3D Gaussian Splatting (1 citation)
- A Step to Decouple Optimization in 3DGS
- Augmented Radiance Field: A General Framework for Enhanced Gaussian Splatting
- CLoD-GS: Continuous Level-of-Detail via 3D Gaussian Splatting
- CryoSplat: Gaussian Splatting for Cryo-EM Homogeneous Reconstruction
- CylinderSplat: 3D Gaussian Splatting with Cylindrical Triplanes for Panoramic Novel View Synthesis
- D$^2$GS: Depth-and-Density Guided Gaussian Splatting for Stable and Accurate Sparse-View Reconstruction
- Distractor-free Generalizable 3D Gaussian Splatting
- FieryGS: In-the-Wild Fire Synthesis with Physics-Integrated Gaussian Splatting
- Flash-Mono: Feed-Forward Accelerated Gaussian Splatting Monocular SLAM
and 54 more in the interactive map.
3D generation · 62 papers
- Generative Human Geometry Distribution (1 citation)
- 3D Aware Region Prompted Vision Language Model
- Aligned Novel View Image and Geometry Synthesis via Cross-modal Attention Instillation
- Articulation in Motion: Prior-free Part Mobility Analysis for Articulated Objects By Dynamic-Static Disentanglement
- ArtVIP: Articulated Digital Assets of Visual Realism, Modular Interaction, and Physical Fidelity for Robot Learning
- AssetFormer: Modular 3D Assets Generation with Autoregressive Transformer
- CTRL&SHIFT: High-quality Geometry-Aware Object Manipulation in Visual Generation
- DiffWind: Physics-Informed Differentiable Modeling of Wind-Driven Object Dynamics
- DreamCS: Geometry-Aware Text-to-3D Generation with Unpaired 3D Reward Supervision
- EgoDex: Learning Dexterous Manipulation from Large-Scale Egocentric Video
- EgoHandICL: Egocentric 3D Hand Reconstruction with In-Context Learning
- FlashWorld: High-quality 3D Scene Generation within Seconds
and 50 more in the interactive map.
Face and shape · 27 papers
- Topology-Preserved Auto-regressive Mesh Generation in the Manner of Weaving Silk (1 citation)
- A Resolution-Agnostic Geometric Transformer for Chromosome Modeling Using Inertial Frame
- Condition Matters in Full-head 3D GANs
- Densemarks: Learning Canonical Embeddings for Human Heads Images via Point Tracks
- DiffTrans: Differentiable Geometry-Materials Decomposition for Reconstructing Transparent Objects
- Direct Reward Fine-Tuning on Poses for Single Image to 3D Human in the Wild
- Einstein Fields: A Neural Perspective To Computational General Relativity
- FastGHA: Generalized Few-Shot 3D Gaussian Head Avatars with Real-Time Animation
- Interp3D: Correspondence-aware Interpolation for Generative Textured 3D Morphing
- Interpretable 3D Neural Object Volumes for Robust Conceptual Reasoning
- LiTo: Surface Light Field Tokenization
- Mesh Splatting for End-to-end Multiview Surface Reconstruction
and 15 more in the interactive map.
Depth and stereo · 27 papers
- Depth Anything 3: Recovering the Visual Space from Any Views (2 citations)
- Dens3R: A Foundation Model for 3D Geometry Prediction (1 citation)
- A Scene is Worth a Thousand Features: Feed-Forward Camera Localization from a Collection of Image Features
- CogniMap3D: Cognitive 3D Mapping and Rapid Retrieval
- CroCoDiLight: Repurposing Cross-View Completion Encoders for Relighting
- Depth Anything with Any Prior
- Joint Shadow Generation and Relighting via Light-Geometry Interaction Maps
- Large Depth Completion Model from Sparse Observations
- Loc$^{2}$: Interpretable Cross-View Localization via Depth-Lifted Local Feature Matching
- Monocular Normal Estimation via Shading Sequence Estimation
- PatchRefiner V2: Fast and Lightweight Real-Domain High-Resolution Metric Depth Estimation
- Physically-Guided Optical Inversion Enable Non-Contact Side-Channel Attack on Isolated Screens
and 15 more in the interactive map.
Human pose estimation · 20 papers
- BigMaQ: A Big Macaque Motion and Animation Dataset Bridging Image and 3D Pose Representations (1 citation)
- Reliable Evaluation of MRI Motion Correction: Dataset and Insights (1 citation)
- From Sparse to Dense: Spatio-Temporal Fusion for Multi-View 3D Human Pose Estimation with DenseWarper
- GenCape: Structure-Inductive Generative Modeling for Category-Agnostic Pose Estimation
- HUMOF: Human Motion Forecasting in Interactive Social Scenes
- Joint Optimization for 4D Human-Scene Reconstruction in the Wild
- Learning Part-Aware Dense 3D Feature Field For Generalizable Articulated Object Manipulation
- Pose Prior Learner: Unsupervised Categorical Prior Learning for Pose Estimation
- QuaMo: Quaternion Motions for Vision-based 3D Human Kinematics Capture
- RECON: Robust symmetry discovery via Explicit Canonical Orientation Normalization
- True Self-Supervised Novel View Synthesis is Transferable
- Zero-shot Human Pose Estimation using Diffusion-based Inverse solvers
and 8 more in the interactive map.
NeRF and radiance fields · 6 papers
- Dynamic Novel View Synthesis in High Dynamic Range
- HDR-NSFF: High Dynamic Range Neural Scene Flow Fields
- PEERING INTO THE UNKNOWN: ACTIVE VIEW SELECTION WITH NEURAL UNCERTAINTY MAPS FOR 3D RECONSTRUCTION
- To View Transform or Not to View Transform: NeRF-based Pre-training Perspective
- Weight Space Representation Learning on Diverse NeRF Architectures
- Nef-Net v2: Adapting Electrocardio Panorama in the wild