Alibaba-MIIL / ASL

Official Pytorch Implementation of: "Asymmetric Loss For Multi-Label Classification"(ICCV, 2021) paper
MIT License
732 stars 102 forks source link

我用了asl加上resnet101,只能达到82.1,论文里面说可以超过84。有朋友有具体的实现细节吗? #47

Closed starmemda closed 3 years ago

sorrowyn commented 2 years ago

目前我的情况比你好一些,最好的结果是82.9。