search

junyanz / CycleGAN

Software that can generate photos from paintings, turn horses into zebras, perform style transfer, and more.
Other
12.22k stars 1.94k forks source link

unpaired images #126

Open hala3 opened 4 years ago

hala3 commented 4 years ago

i use cycleGAN with unpaired images do you think G_los =0.204 , G_Y_loss =0.201 , F_loss=0.8102 and D_X_loss=3.28e-05 is good or not and how i know if my model is in the good way or not

junyanz commented 4 years ago

Your D_X_loss might be too small. It indicates that the discriminator is winning easily. The loss in GANs training might not be as informative as classification tasks. I recommend that you look at the visual results directly.

kiyaaabdian commented 2 years ago

I used semisupervised cyclegan . the losses are DA = 0.007 and GA = 0.8 .do you think the model is failed? if my model is failed what should i do ?

frfdnpg commented 2 years ago

In the end I have not used cyclegan, I have used srgan

El dom, 3 jul 2022 a las 6:11, kiyaaabdian @.***>) escribió:

I used semisupervised cyclegan . the losses are DA = 0.007 and GA = 0.8 .do you think the model is failed? if my model is failed what should i do ?

— Reply to this email directly, view it on GitHub https://github.com/junyanz/CycleGAN/issues/126#issuecomment-1173007882, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABYSCNXAWATTHIRLX7ICXVTVSEHGXANCNFSM4JLT5NSQ . You are receiving this because you are subscribed to this thread.Message ID: @.***>