search

VITA-Group / EnlightenGAN

[IEEE TIP] "EnlightenGAN: Deep Light Enhancement without Paired Supervision" by Yifan Jiang, Xinyu Gong, Ding Liu, Yu Cheng, Chen Fang, Xiaohui Shen, Jianchao Yang, Pan Zhou, Zhangyang Wang
Other
918 stars 200 forks source link

About training loss #82

Closed UtopiaHu closed 3 years ago

UtopiaHu commented 3 years ago

Hi, Thank you for your work. I have a question. Can we get some information from the loss curve during training? What is the normal loss curve like during training? How can I tell the training is not convergent according to the loss curve?

yifanjiang19 commented 3 years ago

The VGG loss is increasing while the adversarial loss is decreasing during training stage. There is no typical signal to show whether a GAN model converges.

UtopiaHu commented 3 years ago

So, 'The VGG loss is increasing while the adversarial loss is decreasing during training stage. ' means the model is not convergent ?

yifanjiang19 commented 3 years ago

on this case and dataset it is not