Get Started
Home
Topics
Search
Library
Research questionHow can we complete unseen 3D scene regions from sparse, unconstrained views without dense 3D supervision?Sparse views leave parts of a scene's geometry and appearance unobserved, while dense 3D representations are expensive and large-scale 3D supervision is limited.
AI
Computer Vision
Image & Video Processing
Machine Learning
Research Paper
Technology
Latest papersRecent research connected to this question, newest first.Sparse auto-regressive modeling for scene generation from multi-view imagesThe source presents SPAR3S, which learns a sparse voxel-aligned 3D latent space from multi-view images using photometric supervision through differentiable 3D Gaussian Splatting. A masked autoregressive transformer predicts missing voxel support and latent tokens. Evidence covers synthetic indoor scenes and RealEstate10k, including novel-view quality.research paper · Sep 3, 2026
Related questions
How can sparse-view 3D scene representations avoid cross-view artifacts without costly per-scene reconstruction?How can multi-view 3D reconstruction predict precise depth while preserving controllable uncertainty over plausible shapes?How can sparse SDF observations constrain a frozen INR’s latent code to complete plausible unseen 3D geometry?How can pretrained 3D scene representations infer depth from unseen viewpoints without task-specific reconstruction training?