KITTI-360 is a large-scale dataset with rich sensory information and complete annotations. It includes 320,000 images and 100,000 laser scan files generated over a driving distance of 73.7 kilometers. Based on the KITTI dataset, KITTI-360 provides more comprehensive 2D and 3D semantic/instance annotations, richer 360° perception information (fisheye images and push-broom laser scans), very accurate and geo-located vehicle and camera positions, and a series of new and challenging benchmarks.