StyleGAN Landscapes Interpolation
ฝัง
- เผยแพร่เมื่อ 22 ก.ย. 2024
- I recreated NVidia's StyleGAN in Keras, find the code here:
github.com/man...
I trained it to generate images of beautiful landscapes, and this is what it's coming up with.
Technical details:
This is linear interpolations between style and noise samples taken every 50 frames (Approx. every 3 seconds).
The model does not use growth or mixing regularities.
It DOES use:
- constant "beginning" of model
- style vector input
- noise image input (with learned scaling)
- Adaptive Instance Normalization (AdaIN)
- Bilinear Upsampling
The final bit causes problems:
if __name__ == "__main__":
model = WGAN(lr = 0.0001, silent = False)
while(True):
model.train()
see below
hi
how to make this video?
ValueError: `tf.compat.v1.keras` Optimizer () is not supported when eager execution is enabled. Use a `tf.keras` Optimizer instead, or disable eager execution.