sample_real_and_fake_latents

[baopmessi]

Thanks for your great work! Sorry for my english
But line 428 in coach.py. It should add latent_avg of style gan ?
It's right or my mistake ?

[omertov]

Hi @baopmessi!
Indeed, when training with the --start_from_latent_avg flag the discriminator should receive the code with the addition of the average latent code.
From my experiments the current version of the repo produces the desired editing results, but I will run an experiment with the addition of the average latents and update the repo afterwards.

Thanks for pointing that out!

[baopmessi]

Looking forward to the results from you soon ! . By the way, results in your paper without addition of the average latent code ?