Improving tracking with a tracklet associator. (arXiv:2204.10677v1 [cs.CV])

Multiple object tracking (MOT) is a task in computer vision that aims to
detect the position of various objects in videos and to associate them to a
unique identity. We propose an approach based on Constraint Programming (CP)
whose goal is to be grafted to any existing tracker in order to improve its
object association results. We developed a modular algorithm divided into three
independent phases. The first phase consists in recovering the tracklets
provided by a base tracker and …

arxiv cv tracking