SceneRig: An Agentic System for Simulation-Ready 3D Scene Reconstruction from Single Images
We present SceneRig, an agentic system that turns a single image into an editable, simulation-ready scene.
ProofPaper ↗
Key points
- Simulation is a powerful tool for robotics, but building diverse, simulation-ready scenes at scale remains difficult.
- Reconstructing scenes from single images is a promising alternative, but visual similarity alone is insufficient: objects must also be recovered, accurately placed, and physically grounded.
- SceneRig combines geometric, visual, and physical tools to reconstruct a faithful 3D scene.
- Furthermore, SceneRig is effective for downstream robotics: its scenes predict real-robot policy outcomes more accurately than the baseline method and more reliably support replay of recorded manipulations without trajectory adaptation.
Sources (1)
- [1]SceneRig: An Agentic System for Simulation-Ready 3D Scene Reconstruction from Single ImagesarXiv (AI, ML, NLP, CV, robotics, multi-agent) · Oct 20, 05:39 PM
We present SceneRig, an agentic system that turns a single image into an editable, simulation-ready scene.
Simulation is a powerful tool for robotics, but building diverse, simulation-ready scenes at scale remains difficult.
Extractive summary: sentences quoted from the sources.
Before this
- Oct 20, 2026How to train MOSS-Image LoRAs locally (8 GB VRAM): settings, tips and workflow
- Oct 19, 2026I built a free, local AI tool that works with PhotoCraft
- Oct 19, 2026Structured outputs for Anthropic Claude models on Amazon Bedrock
- Oct 19, 2026How to continue a GitHub Copilot app session in VS Code
- Oct 19, 2026Push-based KV transfer to improve TTFT and TPOT for disaggregated inference in vLLM
- Oct 19, 2026OV-PEAR: Observation-Verified Pose-Scale Estimation and Reconstruction for Unknown Industrial Manipulation Objects