Our proposed HOMATracker comprises the following steps: (A) Object detection: A detector is used to locate the objects in each frame; (B) Instance-level appearance feature extraction: We involve a ...