Get Started
Home
Topics
Search
Library
Research questionHow can pretrained 3D scene representations infer depth from unseen viewpoints without task-specific reconstruction training?Unseen viewpoints require estimating surfaces that are absent from the directly observed view. The difficulty is determining whether internal representations from pretrained 3D models contain enough scene knowledge to support this inference.
AI
Computer Vision
Diffusion Models
Evaluation & Benchmarks
Image & Video Processing
Machine Learning
Research Paper
Technology
Latest papersRecent research connected to this question, newest first.Zero-Shot Novel Depth Synthesis Using 3D Foundation Models Scene RepresentationsThe source studies depth and pointmap prediction for new views from internal representations of 3D foundation models, including VGGT. It investigates decoding hidden surfaces and uses latent diffusion over those representations; reported results show realistic depth maps across multiple datasets.research paper · Sep 3, 2026
Related questions
How can multi-view 3D reconstruction predict precise depth while preserving controllable uncertainty over plausible shapes?How can we complete unseen 3D scene regions from sparse, unconstrained views without dense 3D supervision?How can novel-view synthesis represent continuous 3D scenes with less capacity and optimization than NeRFs?How can surgical vision models learn scene geometry during pretraining while keeping inference RGB-only?