ICLR 2024 accepted papers on 3D vision
74 of the 2,259 papers accepted at ICLR 2024 (3.3%) are on 3D vision, compared with 2.9% at ICLR 2023. They fall into 6 topics.
See ICLR 2024 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 · 31 papers
- Magic123: One Image to High-Quality 3D Object Generation Using Both 2D and 3D Diffusion Priors (74 citations)
- MVDream: Multi-view Diffusion for 3D Generation (72 citations)
- Let 2D Diffusion Model Know 3D-Consistency for Robust Text-to-3D Generation (34 citations)
- Instant3D: Fast Text-to-3D with Sparse-view Generation and Large Reconstruction Model (32 citations)
- LRM: Large Reconstruction Model for Single Image to 3D (31 citations)
- DreamCraft3D: Hierarchical 3D Generation with Bootstrapped Diffusion Prior (25 citations)
- EmerNeRF: Emergent Spatial-Temporal Scene Decomposition via Self-Supervision (17 citations)
- MagicDrive: Street View Generation with Diverse 3D Geometry Control (10 citations)
- SEPT: Towards Efficient Scene Representation Learning for Motion Prediction (10 citations)
- Symphony: Symmetry-Equivariant Point-Centered Spherical Harmonics for 3D Molecule Generation (4 citations)
- Large-Vocabulary 3D Diffusion Model with Transformer (3 citations)
- Denoising Diffusion via Image-Based Rendering (2 citations)
and 19 more in the interactive map.
NeRF and radiance fields · 17 papers
- ResFields: Residual Neural Fields for Spatiotemporal Signals (4 citations)
- 3D Reconstruction with Generalizable Neural Fields using Scene Priors (2 citations)
- Efficient-3Dim: Learning a Generalizable Single-image Novel-view Synthesizer in One Day (2 citations)
- MovingParts: Motion-based 3D Part Discovery in Dynamic Radiance Field (2 citations)
- H2O-SDF: Two-phase Learning for 3D Indoor Reconstruction using Object Surface Fields (1 citation)
- USB-NeRF: Unrolling Shutter Bundle Adjusted Neural Radiance Fields (1 citation)
- Coordinate-Aware Modulation for Neural Fields
- InsertNeRF: Instilling Generalizability into NeRF with HyperNet Modules
- Learning Robust Generalizable Radiance Field with Visibility and Feature Augmented Point Representation
- Mask-Based Modeling for Neural Radiance Fields
- Neural Field Classifiers via Target Encoding and Classification Loss
- Neural Processing of Tri-Plane Hybrid Neural Fields
and 5 more in the interactive map.
Face and shape · 12 papers
- Continuous Field Reconstruction from Sparse Observations with Implicit Neural Networks (6 citations)
- Sin3DM: Learning a Diffusion Model from a Single 3D Textured Shape (6 citations)
- GPAvatar: Generalizable and Precise Head Avatar from Image(s) (5 citations)
- TUVF: Learning Generalizable Texture UV Radiance Fields (2 citations)
- Ghost on the Shell: An Expressive Representation of General 3D Shapes (1 citation)
- Learning Implicit Representation for Reconstructing Articulated Objects (1 citation)
- TapMo: Shape-aware Motion Generation of Skeleton-free Characters (1 citation)
- $\texttt{NAISR}$: A 3D Neural Additive Model for Interpretable Shape Representation
- An Intuitive Multi-Frequency Feature Representation for SO(3)-Equivariant Networks
- GenCorres: Consistent Shape Matching via Coupled Implicit-Explicit Shape Generative Models
- Pose Modulated Avatars from Video
- Pre-training Sequence, Structure, and Surface Features for Comprehensive Protein Representation Learning
Human pose estimation · 6 papers
- PF-LRM: Pose-Free Large Reconstruction Model for Joint Pose and Shape Prediction (7 citations)
- Cameras as Rays: Pose Estimation via Ray Diffusion (3 citations)
- Towards Category Unification of 3D Single Object Tracking on Point Clouds (3 citations)
- 3D-Aware Hypothesis & Verification for Generalizable Relative Object Pose Estimation (2 citations)
- SparseDFF: Sparse-View Feature Distillation for One-Shot Dexterous Manipulation (2 citations)
- Social-Transmotion: Promptable Human Trajectory Prediction (1 citation)
Depth and stereo · 6 papers
- MVSFormer++: Revealing the Devil in Transformer's Details for Multi-View Stereo (5 citations)
- Self-Supervised High Dynamic Range Imaging with Multi-Exposure Images in Dynamic Scenes (2 citations)
- PORF: POSE RESIDUAL FIELD FOR ACCURATE NEURAL SURFACE RECONSTRUCTION (1 citation)
- Multi-task Learning with 3D-Aware Regularization
- Neural SDF Flow for 3D Reconstruction of Dynamic Scenes
- Improved Efficiency Based on Learned Saccade and Continuous Scene Reconstruction From Foveated Visual Sampling