train和test中的指标不同

sanghyun-son / EDSR-PyTorch

PyTorch version of the paper 'Enhanced Deep Residual Networks for Single Image Super-Resolution' (CVPRW 2017)

MIT License

2.41k stars 666 forks source link

Closed wwang0107 closed 1 year ago

wwang0107 commented 1 year ago

当运行train_IMDN文件时，set5的指标比运行test_IMDN时低了很多，两个文件用的是train中最好的checkpoint