STONE is a large-scale multi-modal dataset for off-road 3D traversability prediction, collected by autonomous ground vehicles across four outdoor environments in South Korea. It provides 7,000 keyframes with surround-view imagery from 6 cameras (1904×1200), 128-channel LiDAR scans (230K points), and voxel-level traversability annotations classifying terrain into free, traversable, potentially traversable, and non-traversable regions. Following the… See the full description on the dataset page:
https://huggingface.co/datasets/Voxel51/STONE.
