search

ZQPei / deep_sort_pytorch

MOT using deepsort and yolov3 with pytorch
MIT License
2.82k stars 726 forks source link

about reid model #81

Open tangsipeng opened 4 years ago

tangsipeng commented 4 years ago

i test the reid model 'ckpt.t7' on Market1501 but only got mAP: 24.8% , can you share your test result ? @ZQPei

tangsipeng commented 4 years ago

i fix some bug on my code ，here is my new test result on Market1501 by 'ckpt.t7' 2020-01-08 10:28:22,563 reid_baseline.inference INFO: Validation Results 2020-01-08 10:28:22,563 reid_baseline.inference INFO: mAP: 36.8% 2020-01-08 10:28:22,563 reid_baseline.inference INFO: CMC curve, Rank-1 :62.2% 2020-01-08 10:28:22,563 reid_baseline.inference INFO: CMC curve, Rank-5 :80.4% 2020-01-08 10:28:22,563 reid_baseline.inference INFO: CMC curve, Rank-10 :86.2%

ZQPei commented 4 years ago

Great job! Can you create a pull request to merge your work to this repository?

tangsipeng commented 4 years ago

@ZQPei
now，I am confused that i chang a strong reid model which can get mAP: 85.7% at Market1501，but did not get any progress，so i wonder that my test result is right or not？

mAP: 36.8% is ok？

antoniolq commented 4 years ago

hello，I am a freshman here. I hava a question about the differences between yolov3 and reid? Can I use yolov3 to extract the features instead of reid model?

CompliceU commented 4 years ago

@tangsipeng 你好，不知道可以请教一下你使用那个strong reid的问题吗

tenghehan commented 3 years ago

@ZQPei now，I am confused that i chang a strong reid model which can get mAP: 85.7% at Market1501，but did not get any progress，so i wonder that my test result is right or not？

mAP: 36.8% is ok？

Hello! I have the same result as you. I replace the ckpt.t7 with some stronger(trained on Market1501 or DukeMTMC) reid model weights but I didn't get any progress on the result of tracking.

sainivedh commented 3 years ago

@tenghehan What did you do next ?. I'm trying with ckpt.t7 but not getting good results. Which Re-ID model should I use

tangsipeng commented 3 years ago

@tenghehan What did you do next ?. I'm trying with ckpt.t7 but not getting good results. Which Re-ID model should I use

when you test on mot task ,you should use a reid model trained on mot data

a865550942 commented 2 years ago

i fix some bug on my code ，here is my new test result on Market1501 by 'ckpt.t7' 2020-01-08 10:28:22,563 reid_baseline.inference INFO: Validation Results 2020-01-08 10:28:22,563 reid_baseline.inference INFO: mAP: 36.8% 2020-01-08 10:28:22,563 reid_baseline.inference INFO: CMC curve, Rank-1 :62.2% 2020-01-08 10:28:22,563 reid_baseline.inference INFO: CMC curve, Rank-5 :80.4% 2020-01-08 10:28:22,563 reid_baseline.inference INFO: CMC curve, Rank-10 :86.2%

Hello, could you tell me how can i get the indicators such as Rank1 and Rank5？