djidje / d-mmd Goto Github PK

我仔细的阅读了一下您公开的代码，发现你在源域和目标域上使用的都是RandomIdentitySampler，所以您在计算MMD loss时直接将一个instances作为同一身份处理，而论文中提到的是将同一tracklet中的帧作为WC样本，将来自不同tracklet的帧作为BC样本，是否有问题？期待您的解惑

duke迁移到market的结果：
mAP: 47.4%
CMC curve
Rank-1 : 68.7%
Rank-5 : 85.7%
Rank-10 : 91.0%
Rank-20 : 94.4%

Hi, thanks for sharing your code.
I got the following error when I tried 'python setup.py develop' at the installation step:
File "setup.py", line 47, in
version=find_version(),
File "setup.py", line 18, in find_version
return locals()['version']
KeyError: 'version'
Could you give me some advice about that?

Besides, when I set up the environment for 'deep-person-reid', everything goes well.

Thanks for your work and releasing your code . after reading your code , I have a question . Did you regard images of same person as a tracklet in target domain .