The classifier judged this item not relevant to AI security, so it is left out of lists, feeds, the API and dataset releases. If that is wrong, report a correction on this page.
InfoResearchPeer-reviewed
Robust 3D Semantic Occupancy Prediction With Calibration-Free Spatial Transformation
- Published
- Record updated
Summary
Researchers present REO, a Robust and Efficient 3D semantic occupancy prediction scheme for autonomous driving that fuses multi-camera and LiDAR data. REO replaces sensor-calibration-dependent 2D-to-3D transformation with a calibration-free spatial transformation based on vanilla attention, and it adds a multi-task training framework and a query-based prediction scheme. On OpenOccupancy, REO achieves a 19.8× speedup over Co-Occ with a 1.1% improvement in geometry IoU.