Autoencoder model weights
Posted: Thu Oct 15, 2020 4:48 pm
When I use "python faceswap.py train (...)", are the model weights randomly initialized? Or are they pre-trained? If so, in which dataset? Imagenet?
The place to discuss Faceswap and Deepfakes
https://forum.faceswap.dev/
When I use "python faceswap.py train (...)", are the model weights randomly initialized? Or are they pre-trained? If so, in which dataset? Imagenet?
They are not pre-trained.
There are various options for initialization (ICNR and Convolutional-Aware) which can be used, otherwise the are generally initialized with He Uniform distribution.