The dataset contains 474 video clips from 215 crowded scenes. Each clip with the extracted trajectories bygKLT trackeris preprocessed by deleting short trajectories, stationary points, and some errors. It contains video name, video length, video size, video source, video t0 (the frame for group detection evaluation), group detection (300 group detection used in our cvpr paper), video_gt (video classes groundtruth), and scene number.These data can only be used for academic research purposes.Related publication:Jing Shao, Chen Change Loy, and Xiaogang Wang. "Scene-Independent Group Profiling in Crowd." in Proceedings of IEEE Conference on Computer Vision and Pattern Recognition (CVPR), 2014. (PDF)