search

grey-eye / talking-heads

Our implementation of "Few-Shot Adversarial Learning of Realistic Neural Talking Head Models" (Egor Zakharov et al.)
GNU General Public License v3.0
593 stars 109 forks source link

Discriminator Working #45

Open atharvasune opened 4 years ago

atharvasune commented 4 years ago

Hi, great work guys! I am a bit curious about the forward method of Discriminator. Why are we returning a list of outputs of Convlayers ? I did not understand the reason behind it. Could someone please explain it to me. :sweat_smile:

nwatab commented 4 years ago

It is because of feature matching. It updates hidden layers in discriminator, so that D(x) and D(G(z)) get closer. From the paper,

a feature matching term [40], which essentially is a perceptual similarity measure, computed using discriminator (it helps with the stability of the training):