Tracking Pedestrian Heads in Dense Crowd

Tracking humans in crowded video sequences is an important constituent ofvisual scene understanding. Increasing crowd density challenges visibility ofhumans, limiting the scalability of existing pedestrian trackers to highercrowd densities. For that reason, we propose to revitalize head tracking withCrowd of Heads Dataset (CroHD), consisting of 9 sequences of 11,463 frames withover 2,276,838 heads and 5,230 tracks annotated in diverse scenes. Forevaluation, we proposed a new metric, IDEucl, to measure an algorithm'sefficacy in preserving a unique identity for the longest stretch in imagecoordinate space, thus building a correspondence between pedestrian crowdmotion and the performance of a tracking algorithm. Moreover, we also propose anew head detector, HeadHunter, which is designed for small head detection incrowded scenes. We extend HeadHunter with a Particle Filter and a colorhistogram based re-identification module for head tracking. To establish thisas a strong baseline, we compare our tracker with existing state-of-the-artpedestrian trackers on CroHD and demonstrate superiority, especially inidentity preserving tracking metrics. With a light-weight head detector and atracker which is efficient at identity preservation, we believe ourcontributions will serve useful in advancement of pedestrian tracking in densecrowds.