STCrowd
收藏资源简介:
STCrowd是由上海科技大学等机构创建的大规模多模态数据集,专注于拥挤场景中的行人感知。该数据集包含219,000个行人实例,平均每帧20人,具有不同程度的遮挡。数据集提供了同步的激光雷达点云和相机图像以及相应的3D标签和联合ID,适用于多种任务,包括仅使用激光雷达、仅使用图像和传感器融合的行人检测和跟踪。数据集的创建旨在解决现有数据集在拥挤场景下行人感知能力的不足,特别是在3D空间中准确检测和跟踪行人的挑战。STCrowd通过提供高密度分布的行人数据,为开发和测试更有效的行人感知方法提供了平台。
STCrowd is a large-scale multimodal dataset developed by ShanghaiTech University and other institutions, focusing on pedestrian perception in crowded scenarios. This dataset contains 219,000 pedestrian instances, with an average of 20 pedestrians per frame, and exhibits varying degrees of occlusion. It provides synchronized LiDAR point clouds, camera images, as well as corresponding 3D annotations and joint IDs, supporting a wide range of tasks including pedestrian detection and tracking via LiDAR-only, image-only, and sensor fusion modalities. The dataset was created to address the limitations of existing datasets in pedestrian perception within crowded scenarios, especially the challenges of accurate pedestrian detection and tracking in 3D space. By offering high-density pedestrian distribution data, STCrowd provides a platform for developing and validating more effective pedestrian perception methods.




