The first self-supervised, domain-agnostic model for LiDAR ground segmentation, eliminating the need for per-sensor manual labeling.
March 31, 2026
Original Paper
TerraSeg: Self-Supervised Ground Segmentation for Any LiDAR
arXiv · 2603.27344
The Takeaway
By training on a standardized dataset of 22 million scans across 15 sensor types, it provides a 'universal' ground segmenter that practitioners can deploy on any LiDAR hardware with state-of-the-art zero-shot performance.
From the abstract
LiDAR perception is fundamental to robotics, enabling machines to understand their environment in 3D. A crucial task for LiDAR-based scene understanding and navigation is ground segmentation. However, existing methods are either handcrafted for specific sensor configurations or rely on costly per-point manual labels, severely limiting their generalization and scalability. To overcome this, we introduce TerraSeg, the first self-supervised, domain-agnostic model for LiDAR ground segmentation. We t