RandomErasing and reproducing the result

[bmiftah]

@HRanWang Thank you for sharing this and appreciated much.

I try to reproduce the result on Market-1501 following the recommended setting and parameters , I wasnt able to get near the result in the paper. I keep the default learning rate and weight as in the code (lr = 0.01 and weight decat 5e-04).. and I use the same parameters list given here to run .. I noted in the paper Random erasing with probability of 0.5 is used but I dont see this in the code .. Has anybody used RandomErasing , anybody was able to reproduce near or same result .. .Thank you