NVIDIA Releases Foundation4D, a Unified 4D Perception Model That Ditches Lidar
jonLorraine9 · x · 2026-10-30
NVIDIA released Foundation4D, a unified 4D perception foundation model for understanding and reconstructing driving scenes. From camera input alone it generates depth, camera poses, 3D objects, occupancy, and Gaussian splats — camera-based annotations for training plus Omniverse NuRec reconstruction, simplifying dataset preparation without lidar.
Team member lealtaixe highlighted the appeal of one model with shared video features supporting multiple scene-understanding outputs.
More from Embodied
- FACT learns dexterous robot manipulation from human video plus force signals — jiajunwu_cs · 2026-10-30
- Gemini 4 Argon tops 4 robotics tasks on RLE-Bench, ranks #2 behind GPT-6 Astra — daibond_alpha · 2026-10-30
- NVIDIA Opens Cosmos World Model Training Stack, Adds RL Support for Robotics — NVIDIA Developer · 2026-10-30
- SF startup pays $20 an hour to walk robot dogs for AI training data — Medical-Decision-125 · 2026-10-30
- Robot contact prediction idea: mask body patches by imminent-contact likelihood — breadli428 · 2026-10-30
- OpenArm 1.0 autonomously cleans up a spill using real-time demo-to-policy pipeline — stepjamUK · 2026-10-30