3DV 2026

164 papers
The 2026 edition of the International Conference on 3D Vision (3DV) featured 147 papers, reflecting diverse developments in three-dimensional reconstruction and scene understanding. A primary focus of the research involved 3D Gaussian Splatting, with studies addressing frequency-aware decomposition and the application of diffusion models to enhance reconstruction quality. The integration of Large Language Models (LLMs) into 3D workflows was also prominent, particularly in the context of automated interior design and layout generation. Additional research themes included the synthesis of interactable digital twins from casual video and the recovery of human motion under occlusions. The year’s proceedings featured contributions from authors such as Christian Theobalt, Daniel Cremers, Yebin Liu, and Vladislav Golyanik. Notable papers explored goal-conditioned multimodal transformers for object trajectory synthesis and instruction-guided semantic navigation. Furthermore, advancements in sensing were represented by work on induced vibration for event-based systems, indicating a trend toward more resilient and instruction-aware spatial computing.

Top Research Directions

Research directions most represented by 3DV 2026 papers.

Aligning Generative Media with Human Perceptual Preferences 3
Hybrid Representations for High-Fidelity 3D Reconstruction 3
3D Gaussian Splatting for Photorealistic Scene Representation 3
Eliminating Per-Scene Optimization with Feed-Forward 3D Reconstruction 3
Resolving Shape Radiance Ambiguity in Sparse View Reconstruction 3
Real Time View Synthesis with 3D Gaussian Splatting 2
Generative Simulation for Scaling Embodied AI 2
Compositional Generative Models for Multi Object 3D Scenes 2
Error Propagation in Monocular Depth Priors 2
Pose Free 3D Reconstruction and Novel View Synthesis 2
Aligning Video Generation Models with Physical Laws 2
Sharp 3D Reconstruction from Motion Blurred Imagery 2
Cascaded Error Propagation in 2D to 3D Lifting 2
Overcoming the Static Scene Assumption in Computer Vision 2
Fusing 2D Semantic Features with 3D Geometric Representations 2

Top Authors

AuthorPapers
Christian Theobalt7
Daniel Cremers6
Florian Bernard4
Nassir Navab4
Vladislav Golyanik4
Yebin Liu4
Ethan Weber3
Federico Tombari3
Kris Kitani3
Kunyi Li3
Marc Habermann3
Michael J. Black3
Ping Tan3
Sen Wang3
Siyun Liang3
Stefano Gasperini3
Thabo Beeler3
Adam Kortylewski2
Anand Bhattad2
Andrea Tagliasacchi2

Papers

Title
PromptVFX: Text-Driven Fields for Open-World 3D Gaussian Animation
Proxy-Free Gaussian Splats Deformation with Splat-Based Surface Estimation
Quantum Multiple Rotation Averaging
RaCo: Ranking and Covariance for Practical Learned Keypoints
RadarSim: Simulating Single-Chip Radar Via Multimodal Neural Fields
Radially Distorted Homographies, Revisited
Reconstructing Hand-Held Objects in 3D from Images and Videos
Recovering 3D Shapes from Ultra-Fast Motion-Blurred Images
Reflect3r: Single-View 3D Stereo Reconstruction Aided by Mirror Reflections
Revisiting Depth Representations for Feed-Forward 3D Gaussian Splatting
RGS-DR: Deferred Reflections and Residual Shading in 2D Gaussian Splatting
SAIL-Recon: Large SfM by Augmenting Scene Regression with Localization
SAMa: Material-Aware 3D Selection and Segmentation
SAT-SKYLINES: 3D Building Generation from Satellite Imagery and Coarse Geometric Priors
SceneGen: Single-Image 3D Scene Generation in One Feedforward Pass
SCENIC: Scene-Aware Semantic Navigation with Instruction-Guided Control
SCIGaussian-D: Dynamic Scene Reconstruction from a Single Snapshot Compressive Image
Seeing Through Clutter Structured 3D Scene Reconstruction via Iterative Object Removal
Seeing Through Clutter: Structured 3D Scene Reconstruction via Iterative Object Removal
Self-Supervised Implicit Attention Priors for Point Cloud Reconstruction
Semantic-Free Procedural 3D Shapes Are Surprisingly Good Teachers
SemLayoutDiff: Semantic Layout Generation with Diffusion Model for Indoor Scene Synthesis
SH-SAS: An Implicit Neural Representation for Complex Spherical-Harmonic Scattering Fields for 3D Synthetic Aperture Sonar
SmokeSeer: 3D Gaussian Splatting for Smoke Removal and Scene Reconstruction
SNAP: Towards Segmenting Anything in Any Point Cloud
SocialGen: Modeling Multi-Human Social Interaction with Language Models
SPATIALGEN: Layout-Guided 3D Indoor Scene Generation
Splat-based Gradient-domain Fusion for Seamless View Transition
Splatography: Sparse multi-view dynamic Gaussian Splatting for film-making challenges
Splatography: Sparse Multi-View Dynamic Gaussian Splatting for Filmmaking Challenges
SpotLight: Shadow-Guided Object Relighting via Diffusion
Structurally Disentangled Feature Fields Distillation for 3D Understanding and Editing
Structure-Grounded Training Strategies Aid Generalization in Stereo Matching
SuperGSeg: Open-Vocabulary 3D Segmentation with Structured Super-Gaussians
Supervising 3D Talking Head Avatars with Analysis-by-Audio-Synthesis
Symmetry Informative and Agnostic Feature Disentanglement for 3D Shapes
Tesselation GS: Neural Mesh Gaussians for Robust Monocular Reconstruction of Dynamic Objects
Tessellation GS: Neural Mesh Gaussians for Robust Monocular Reconstruction of Dynamic Objects
TexAvatars: Hybrid Texel-3D Representations for Stable Rigging of Photorealistic Gaussian Head Avatars
Text-to-3D Generation Using Jensen-Shannon Score Distillation
TextureSplat: Per-Primitive Texture Mapping for Reflective Gaussian Splatting
Toon3d: Seeing Cartoons From New Perspectives
Too Tiny to See: Hazardous Obstacle Detection Dataset and Evaluation
Towards Enhanced Sparse-View Tomographic Reconstruction Using 3D Gaussian Splatting
TRASE: Tracking-Free 4D Segmentation and Editing
Triangle Splatting for Real-Time Radiance Field Rendering
TrueCity: Real and Simulated Urban Data for Cross-Domain 3D Scene Understanding
TwoSquared: 4D Generation from 2D Image Pairs
Uncertainty-Aware 3D Edge Reconstruction with Difference of Gaussians
Unified Geometry and Color Compression Framework for Point Clouds via Generative Diffusion Priors

« prev Page 3/4 next »