SemanticKITTI
lastmod 2026-11-10created 2026-09-292 edits
SemanticKITTI is a large-scale 3D point cloud dataset that adds point-wise semantic labels to the LiDAR scans of the KITTI dataset for autonomous driving research. It is annotated with many classes such as road, vehicle, pedestrian, and building, and serves as a standard benchmark for tasks like 3D semantic segmentation, panoptic segmentation, and scene completion. Point cloud training frameworks such as Pointcept support it as a default dataset.