Overview
Research area: Computer Vision Authors: Inhee Lee, Sangwon Baik, Sungjoo Kim, Hyeonwoo Kim, Hyunsoo Cha, Hanbyul Joo Published: 2026-06-02 arXiv: 2606.03994
Summary
Reconstructing interactive, simulation-ready 3D scenes from a single image is a critical bottleneck for robotic manipulation. While recent single-image lifters recover plausible per-object shapes, composing them yields scenes that collapse under physical simulation due to interpenetrating, hovering, or sinking objects. Existing physics-aware methods address this strictly as a post-hoc layout correction, leaving the underlying geometric errors unresolved.
To address this, the authors introduce SimuScene, a compositional 3D reconstruction pipeline that puts physics in the loop of shape and layout estimation. Rather than using physics merely for layout cleanup, the system utilizes the physics engine as a diagnostic measurement tool during the generative process itself. By diagnostically simulating reconstructed objects under gravity, SimuScene converts penetration and support failures into quantitative correction signals that drive gravity-axis stretching and amodal shape resampling.
This physics-informed feedback loop mitigates accumulated reconstruction errors, producing stable, simulation-ready compositional 3D scenes. Extensive experiments demonstrate state-of-the-art performance on physical stability and geometric alignment benchmarks.
Key contributions
- Physics integrated into the shape and layout estimation loop, not just post-hoc layout cleanup
- Physics engine used as a diagnostic measurement tool during generation
- Penetration and support failures converted into quantitative correction signals (gravity-axis stretching, amodal shape resampling)
- State-of-the-art results on physical stability and geometric alignment benchmarks