Weevaluatetheperformanceofawidelyusedtrackingby-detectionanddataassociationmulti-targettrackingpipelineappliedtoanactivity-richvideodataset.Incontrasttotraditionalworkonmulti-targetpedestriantrackingwherepeoplearelargelyassumedtobeupright,weuseanactivity-richdatasetthatincludesawideran