Abstract
In this paper, we propose a label propagation frameworkto handle the multiple object tracking (MOT) problem for ageneric object type (cf. pedestrian tracking). Given a tar-get object by an initial bounding box, all objects of the sametype are localized together with their identities. We treat thias a problem of propagating bi-labels, i.e. a binary class label for detection and individual object labels for tracking. To propagate the class label, we adopt clustered Multiple Task Learning (cMTL) while enforcing spatio-temporal consistency and show that this improves the performance when given limited training data. To track objects, we prop-agate labels from trajectories to detections based on affin-ity using appearance, motion, and context. Experiments on public and challenging new sequences show that the proposed method improves over the current state of the art on this task.