AHa-3D turns ordinary indoor videos into editable Blender 3D scenes
ducha_aiki · x · 2026-09-28
The open-source AHa-3D project presents an agentic Real2Sim pipeline that converts an ordinary indoor video into an editable Blender 3D scene. It rebuilds the room from reference geometry plus a library of reusable furniture and materials, estimates human motion, places people in the shared scene with foot-ground and object contact refinement, renders with the original camera and timing, and can generate new actions or export interactive browser demos. The author notes dedicated 3D vision models still beat frontier LLMs alone on temporal consistency, precision and efficiency, though end-to-end internal 3D understanding remains a goal worth chasing.
More from Embodied
- HSImul3R (ECCV 2026): simulation-ready human-scene reconstruction judged by physics, not looks — jiqizhixin · 2026-09-28
- World's largest humanoid robot livestream performance draws crowd of 10,000+ — kernelangus420 · 2026-09-28
- One-arm VR intervention during bimanual DAgger feels like an AI brain chip — neurosp1ke · 2026-09-28
- USTC's VLA-Precision: online RL for VLAs hits 98.3% success on precision chemistry tasks — ustc · 2026-09-28
- Robot training fix: cutting noise injection from 2.5% to 0.25% of joint range removes jerkiness — DominiqueCAPaul · 2026-09-28
- First real-world RL fine-tuning on a VLA: policy turns jerky after ~10 rollouts — DominiqueCAPaul · 2026-09-28