William

I think there are some mistakes

1

After applying Channel Attention Module, maybe it would be better to apply a convolution layer in order to modify the channels to the original value (usually 3 channels), instead of...

None {}/{}===> processing {} errored ! [ WARN:[email protected]] global loadsave.cpp:244 cv::findDecoder imread_('/home/cgy/Chang/image_enhancement/datasets/fiveK/input/a0812-IMG_3125.jpg'): can't open/read file: check file path/integrity Traceback (most recent call last): File "d:\DeepLearning\Essays\ShadowRemoval\Invertible\Underexposed Photo Enhancement2019 CVPR\DeepUPE_pytorch\test.py", line 91,...

refusion.yml和ir-sde.yml的区别？

5

只是将UNet换成NAFNet，并没有使用[Refusion]这篇文章中提到的训练方法？ （不过NAFNet的效果确实好很多）

我用ISTD数据集训练，Loss在1e-4左右，但是测试的时候，输出的是噪声，和原图以及去除阴影之后的图都没有任何相似之处

1

是不是保存的图片不对