Head Tracking 21 Head Tracking Dataset

CroHD stands for Crowd of Heads Dataset, which is a dataset that can be used for head tracking tasks. The dataset consists of 11,463 frames in 9 high-definition resolution sequences, with more than 2,276,838 heads and 5,230 tracks, which are annotated in different scenes.

All sequences in CroHD were shot at 25 fps from an elevated viewpoint. They involve crowded indoor and outdoor scenes, and are shot under varying lighting and environmental conditions.