hello Ziqi, thans for your great work. I have a question about the tracklet score update, in your code:
- the tracklet score is initialized with
self.latest_score = bbox.s
- in the predict step
self.latest_score = self.latest_score * 0.01
- in the update step
self.latest_score = update_info.bbox.s
since your life manager is based on hit instead of score, what's the point of the tracklet score that you use? i tried to modifed the the score decay coefficient for self.latest_score in the predict step, and find that the overall result is worse, but I can't find anything in the code that use the tracklet score. do you know the reason? thank you
hello Ziqi, thans for your great work. I have a question about the tracklet score update, in your code:
self.latest_score = bbox.s
self.latest_score = self.latest_score * 0.01
self.latest_score = update_info.bbox.s
since your life manager is based on hit instead of score, what's the point of the tracklet score that you use? i tried to modifed the the score decay coefficient for self.latest_score in the predict step, and find that the overall result is worse, but I can't find anything in the code that use the tracklet score. do you know the reason? thank you