performance question on mcn-mt

[wangyuan249]

Hi~, I am currently following your research work in cross-domain pedestrian re-identification. I would like to ask a few questions:

1. In the performance table of cuhk03 -> market and cuhk03 -> duke, is there provided the baseline-mmt (mutual mean teacher) performance index as a comparation?
2. How is the migration effect of cuhk03 itself as the target domain? Our current experiments on cuhk as the target domain are very poor. Are there some problems with cuhk's map and top1 evaluation indicators?

[JeremyXSC]

Thanks for your interest in our work. Q1: So far we have not performed the baseline-mmt evaluation on cuhk03 -> market and cuhk03 -> duke respectively, but we are very interested in exploring them in the future. Q2: Probably, I think the main reason for this phenomenon is that the query images in CUHK03 dataset are very challenging, so the test performance on this dataset is generally not satisfied.