Reproducibility issue

[RaviShivam]

I have some problems with reproducing the mAP value reported in your article for the DukeMTMC-reID dataset. Here are the results when I run the standard evaluation code provided by DukeMTMC-reID_evaluation. I used the trained Duke Dense201 model you have provided and here are my results:

single query: mAP = 0.674285, r1 precision = 0.843750

In your article you had reported a mAP of 80.0%. Could you please indicate if I have done anything wrong whilst running your model?