← All resources

Reconstructing Physically Stable 3D Scenes from a Single Image

Auto-discovered

Paper Advanced

Xiaoxuan Ma, Jiashun Wang, Nicolás Ugrinovic, Yehonathan Litman, Kris Kitani

Visit resource ↗

Description

Reconstructing Physically Stable 3D Scenes from a Single Image is a single research paper with a public REST3D project site, code release, and demo videos. REST3D: Reconstructing Physically Stable 3D Scenes from a Single Image presents a framework that builds a gravity-support scene tree, initializes with image-to-3D models, then applies physics-constrained refinement so.

The publication explains the method, training or solve details, and reports comparisons against prior lens models, mesh extractors, or reconstruction baselines.

It targets advanced researchers and engineers in rendering, vision, or geometry who already work with the problem domain. Authors listed in the submission include Xiaoxuan Ma, Jiashun Wang, Nicolás Ugrinovic, Yehonathan Litman, Kris Kitani.

Read the project page or PDF for figures, ablations, and any supplemental video; linked assets are grouped on the same site as the paper listing.

It is useful when you need implementation-oriented context and quantitative claims before reading the full PDF. Link last verified accessible on 2026-09-28.