it seems the 224 lane of detect_and_track.py should be:
txt_str += " %f %f" % (np.abs(track.bbox_history[-1][0] - track.bbox_history[-1][2]) / im0.shape[1], np.abs(track.bbox_history[-1][1] - track.bbox_history[-1][3]) / im0.shape[0])
the first value divided by im0.shape[1].
im0.shape[1] is the width of the picture and track.bbox_history[-1][0] is the x of object.
it seems the 224 lane of detect_and_track.py should be:
txt_str += " %f %f" % (np.abs(track.bbox_history[-1][0] - track.bbox_history[-1][2]) / im0.shape[1], np.abs(track.bbox_history[-1][1] - track.bbox_history[-1][3]) / im0.shape[0])
the first value divided by im0.shape[1]. im0.shape[1] is the width of the picture and track.bbox_history[-1][0] is the x of object.