Voxel51/STONE
收藏资源简介:
STONE是一个用于越野3D可通行性预测的大规模多模态数据集,由自主地面车辆(UGV)在韩国四个户外环境中收集。数据集遵循nuScenes格式,提供7,000个关键帧,包括来自6个摄像头的环绕视图图像(分辨率1904×1200)、128通道LiDAR扫描(约230,400点)以及体素级可通行性注释,将地形分类为自由、可通行、潜在可通行和不可通行区域。此外,数据集包含3D障碍物边界框、自我姿态轨迹和约10 Hz的同步多传感器数据。这个FiftyOne版本是从完整的279个场景集合中分层抽取的35个场景(每个场景200帧),组织为分组样本,每个关键帧有7个切片(对应6个摄像头和1个LiDAR 3D场景)。
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 nuScenes format, the dataset includes 3D obstacle bounding boxes, ego-pose trajectories, and synchronized multi-sensor data at ~10 Hz. This FiftyOne version contains a stratified sample of 35 scenes (200 frames each) from the full 279-scene collection, organized as grouped samples with 7 slices per keyframe (6 cameras + 1 LiDAR 3D scene).




