Train result
Opened this issue · 1 comments
Kim-Z commented
excuse me, I train the model by myself base on the default veri.yml setting: backbone ResNeXt101 ibn a, multi-head, CE and SupCon losses, MALW and imagenet pretrained, but only reach 70% mAP, Is my parameter not set correctly?
jinxinzheok commented
excuse me,I meet some questions when i reproduce the code. can i ask you some questions my email 1119157406@qq.com