Market 1501
Resnet50 : batchsize=32 : Rank1 =95.81 , mAP=88.28
- RK (Re-Ranking): Rank1 =96.25 , mAP=94.31
If you want a higher accuracy, you may modify batchsize or use Resnet101.
DukeMTMC-REID
Resnet50: batchsize =32 : Rank1 =88.7 , mAP=77.9
pytorch = 0.4