SemanticKITTI extends the capabilities of the original KITTI dataset by adding a semantic layer to the LiDAR point clouds. This enhancement opens new avenues for research and development in 3D scene understanding and LiDAR-based perception for autonomous vehicles.

Rich Semantic Labeling: Each point in the LiDAR scans is annotated with semantic labels, providing detailed insights into the environment. This granularity is key for tasks like semantic segmentation and object classification.

Dynamic Scene Understanding: With its focus on dynamic environments, SemanticKITTI is particularly suited for understanding and predicting the behavior of various elements in a driving scenario, from pedestrians to other vehicles.

Large-Scale LiDAR Data: The dataset’s extensive collection of LiDAR frames, annotated with semantic information, offers a unique resource for developing and refining LiDAR-based perception algorithms, crucial for fully autonomous driving systems.

Semantic + panoptic segmentation (SemanticKITTI): 11 labeled sequences, 28 categories
Information + label format (email required)

You can preview and copy a sample of this dataset on

Share this dataset