Search papers, labs, and topics across Lattice.
University of Freiburg
5
0
6
35
Ditch the LiDAR: This camera-only method uses pre-built point cloud maps to drastically improve 3D object detection and tracking, rivaling LiDAR-based systems in familiar environments.
Radar and camera don't just see more together, they see *differently* - and this fusion architecture shows exactly how to leverage each sensor's unique strengths for superior 3D object detection.
Expert ordinal comparisons reveal that fusing vision and language in wound representation learning boosts agreement by 5.6% over unimodal foundation models for a rare genetic skin disorder.
Current autonomous driving metrics miss the most dangerous perception errors: effort-based metrics reveal that up to 93% of errors are non-critical and identify the failures that truly matter for safety.
Achieve state-of-the-art 4D scene understanding by representing dynamic environments as a collection of Gaussians that are efficiently splatted onto a voxel grid for panoptic occupancy tracking.
