detectiveli / PSMA

Progressive Sample Mining and Adversarial Learning for One-Shot Person Reidentification

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Comparision experiments EUG v.s. PSMA

HXYNODE opened this issue · comments

Hi,i noticed that the Table1. shows the comparision experiments results related with some previous works in the experiments part of your paper. And the results of the EUG method are listed below (from your paper):
EUG [7] | One-shot | 55.8 72.3 78.4 26.2 | 48.8 63.4 68.4 28.5
however, the experiment results in original EUG paper are followings (EUG):
EUG (p=0.30) | One-shot | 42.77 56.51 67.17 21.12 | 63.82 78.64 87.04 54.57
...
EUG (p=0.05) | One-shot | 62.67 74.94 82.57 42.45 | 72.79 84.18 91.45 63.23
what's the differences betweent the experiment results listed in [7] and yours? and why you didn't compare the higher results EUG (p=0.05) ? is there anything i misunderstand? expect your reply!

commented

Hi,

Could you please offer me the reference where you got the result?

Cause the original paper we referenced is from :
https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8607049

Thanks.

i got the results from https://ieeexplore.ieee.org/document/8578641
actually both of them are from the same reseacher. Also Table II of https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=8607049 shows the results i displayed.

commented

One dataset is from video Reid (MARS), and the other is from image-based Reid (Market1501).

thanks a lot! i got it. do you have performed some experiments on the dataset (DukeMTMC-VideoReID) with your model and what about the results?