search

vincent-thevenin / Realistic-Neural-Talking-Head-Models

My implementation of Few-Shot Adversarial Learning of Realistic Neural Talking Head Models (Egor Zakharov et al.).
GNU General Public License v3.0
828 stars 195 forks source link

Generator architecture doesn't match the one described in paper #50

Closed Jarvisss closed 4 years ago

Jarvisss commented 4 years ago

First, thanks a lot for the sharing of your implementation.

But I notice there is an architecture mismatch in your code and the original paper:

https://github.com/vincent-thevenin/Realistic-Neural-Talking-Head-Models/blob/e43ca9fccc297ca048544722aab1a3ee4e9b59f5/network/model.py#L156-L173

image

you use AdaIN in the decode part of generator, while in the original paper they use IN.

Is this change made for better result?

Jarvisss commented 4 years ago

sorry for the misunderstanding. Your implementation is correct, I'll close this issue