search

ligoudaner377 / font_translator_gan

88 stars 9 forks source link

this performance is from L1loss ？ #3

Open Johnson-yue opened 2 years ago

Johnson-yue commented 2 years ago

hi， thank your share the code. I have an question about loss

did you analysis loss importance for w & w/o L1 loss ?

the result is for blur , if w/o gan loss , the performance maybe the same ?

ligoudaner377 commented 2 years ago

Hi! @Johnson-yue I didn't try "w & w/o L1 loss". https://github.com/ligoudaner377/font_translator_gan/blob/9e1aaf03b3edbacee0023607f60cc4b2a155cc8b/models/font_translator_gan_model.py#L123 But here's a hyperparameter self.opt.lambda_L1, it can control the weight of L1 loss. If you set it to 0 and train the model, basically it has the same effect with "w/o L1 loss" hope this can help you!