Abstract
Multiple object tracking (MOT) in urban traffic aims to produce thetrajectories of the different road users that move across the field of viewwith different directions and speeds and that can have varying appearances andsizes. Occlusions and interactions among the different objects are expected andcommon due to the nature of urban road traffic. In this work, a trackingframework employing classification label information from a deep learningdetection approach is used for associating the different objects, in additionto object position and appearances. We want to investigate the performance of amodern multiclass object detector for the MOT task in traffic scenes. Resultsshow that the object labels improve tracking performance, but that the outputof object detectors are not always reliable.