t0903 / Re-ID

Re-ID

1:prepare.py

2:train_11.py

3:test.py

4:evaluate_gpu.py

Market 1501

        Resnet50 :          batchsize=32 :         Rank1 =95.81    ,    mAP=88.28


        +  RK (Re-Ranking):  Rank1 =96.25    ,     mAP=94.31


       Resnet50(IBN) :    batchsize=32 :         Rank1 =95.8    ,    mAP=89.24

DukeMTMC-REID

       Resnet50:          batchsize =32 :        Rank1 =88.84     ,    mAP=77.95


      Resnet50(IBN):     batchsize =32 :        Rank1 =90.1    ,    mAP=79.7

      pytorch = 0.4

If you want a higher accuracy, you may modify batchsize .