Performance on MSMT17

[crazydemo]

I use the default setting and iters=800 to train spcl_uda, source dataset=msmt17 and target dataset=market1501. I only get map=72.0, rank1=87.5, which is far away from your claimed metrics. Are there any tricks, or some missing points I should focus on?

[yxgeee]

I used iters=400. p.s. iters=800 is only adopted when msmt17 or vehicle datasets are used as the target-domain dataset.

[xmy0916]

@yxgeee 大佬好，有个问题一直比较困惑，为什么market和duke的指标都有七八十那么高，然后msmt17的指标才二三十呢？这是因为数据集有什么问题吗？