search

tadax / glcic

Globally and Locally Consistent Image Completion with TensorFlow
MIT License
135 stars 42 forks source link

Is this GAN? #10

Open takeseijin opened 6 years ago

takeseijin commented 6 years ago

Hi, thank you for your much effort! I think your code has two fatal mistakes. (1. As issue #2, your generator gets no feed back from discriminator. Use joint-loss.)

  1. On your code, there's no path to back-propagate "discriminator loss in joint-loss" (I mean log(1-D) in joint loss) to generator's weights and biases. Because your class Network gets completion and local_completion from outside of the network. (You can confirm this by opt.compute_gradients(model.d_loss,model.g_variables) which returns None instead of gradient tensor.)
675492062 commented 5 years ago

I can't agree more！It’s not GAN here.