Closed jakeelwes closed 5 years ago
Hi @jakeelwes -- for a simpler example, take a look at this file from issue #1 -- it only predicts a single 512-wide vector for everything, so that includes the pose. This is also what StyleGAN does while training.
I'm not originally from an ML background either, I've just been programming for a long time!
Hey @pbaylies thanks for this repo! :) You suggested to predict pose I could train a resnet to predict just the coarse dlatents (Puzer/stylegan-encoder#15).
Would it be something to do with this reshape? https://github.com/pbaylies/stylegan-encoder/blob/b5ddcd7a820146e3b45182789518e2ff499704f3/train_resnet.py#L162
Or changing the size of W? https://github.com/pbaylies/stylegan-encoder/blob/b5ddcd7a820146e3b45182789518e2ff499704f3/train_resnet.py#L55-L60
Thanks for your help, sorry I'm not from a ML background.