Hello! I’m using official pytorch examples repo to train my model. I want to use high resolution images (1500x2000) and spent many hours with trying different values of style-weight and content-weight, but I found that the size of grid of effect is always same
Here is my current result:
You can see, that the grid of effect is very small, so it is trying to reproduce style in small squares. I want to copy bigger parts of these fractals in my image.
How can I control it?
As I understand I can increase dimensions of layers here: https://github.com/pytorch/examples/blob/master/fast_neural_style/neural_style/transformer_net.py
But then my mode has much more bigger size.
Is it good idea or not? Or I can change it somewhere in this code?
Many thanks!