search

CaptainEven / MCMOT

Real time one-stage multi-class & multi-object tracking based on anchor-free detection and ReID
MIT License
383 stars 82 forks source link

[Question] Det vs Track confidence #46

Closed austinmw closed 3 years ago

austinmw commented 3 years ago

I noticed that self.det_thresh = opt.conf_thres. I've seen in some models that they use a lower thresh to continue track compared to init new track, have you happened to see how that compares?

CaptainEven commented 3 years ago

@austinmw It may be a better strategy for some specific data-set, but i'm not sure. A lower confidence thresh impose less rigorous standard for verifying the track. As far as i know, this may help to improve the track recall but will harm the precision, so it's only a balance strategy, do not matters much i think.

CaptainEven commented 3 years ago

@austinmw The code has updated for writing detection score to the results.txt in the 7th(index 6) column.