affinelayer / pix2pix-tensorflow

Tensorflow port of Image-to-Image Translation with Conditional Adversarial Nets https://phillipi.github.io/pix2pix/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Short question about Discriminator Loss

bizcjh1027 opened this issue · comments

image

As you can see above, my discriminator loss fluctuate near the 0.5.

Does that mean training result is quite successful? judging by an optimal probability of discriminator is 1/2(which means discriminator isn't sure about which is fake or real).

Could you give me an advise for it?

I think optimal probability of discriminator 1/2 should be the output of D, not the loss function. So, the corresponding value of loss should be around 0.69