search

Yang-Liu1082 / InvDN

Implementation for the paper: Invertible Denoising Network: A Light Solution for Real Noise Removal (CVPR2021).
Apache License 2.0
185 stars 35 forks source link

About training loss #25

Closed YangGangZhiQi closed 1 year ago

YangGangZhiQi commented 1 year ago

Hi, thanks for this great works! I try to train invdn in my own datasets. I get a strange problem. The l_back_rec is too big, eg 1e+3. I don't know whether it is normal. InvDN_back_rec_loss太大 I would apprecaite it if someone could help me.

YangGangZhiQi commented 1 year ago

Today I figure it out. Because the loss function calculates the sum of x-target, so the results is eCHW。For example, C=3, H=W=144, average error e=0.5, the loss result=0.53144144=3.1104e+4. So the loss results above is normal.

8056 commented 1 year ago

Thank you, could you mind to share the final loss during your trainning ?

sui1999 commented 1 year ago

今天我想通了。因为损失函数计算的是x-target的总和，所以结果是e C H W。比如C=3，H=W=144，平均误差e=0.5，损失结果=0.5 3 144 144=3.1104 e+4。所以上面的损失结果是正常的。

Hi ,May I ask how you configured the code after downloading it? I don't quite understand how my training is particularly slow