NVIDIA's Instant NuRec reconstructs a drivable 3DGS world from driving logs in ~1.5 seconds
rsasaki0109 · x · 2026-10-01
NVIDIA's Spatial Intelligence Lab introduced Instant NuRec, a feed-forward neural reconstruction model that turns a short multi-camera driving log into a fully simulatable layered 3D Gaussian Splatting world in a single forward pass — including static/dynamic 3DGS layers, a sky cubemap, per-camera ISP corrections, and native non-pinhole camera support via 3DGUT.
- Reconstructs 10–20s multi-camera scenes in 1.5 seconds, no per-scene tuning needed
- Beats the strongest baseline by 2.01 dB PSNR on the Waymo Open Dataset
- Deeply integrated into NuRec and compatible with AlpaSim for closed-loop simulation
- Paper, code, and model are released
More from Embodied
- Robotics companies aren't taking 'cheapness' seriously enough, argues founder — hudzah · 2026-10-01
- Blogger predicts humanoid robots will soon run hotel services like laundry — CyberRobooo · 2026-10-01
- GPT-6 Astra robot evals show strong task decisions but weak physical control — Galbot · 2026-10-01
- RoboCoach uses world-model-imagined failures as coaching: 13.3% to 75% success with 150 demos — Tsinghua · 2026-10-01
- GGSD: Game Self-Play Yields Human-Playable Robot Skills That Solve Unseen Tasks Without Retraining — Seungeun Rho · 2026-10-01
- Microsoft Research open-sources Rho, a family of 5B VLAs for adaptable physical AI — chris_j_paxton · 2026-10-01