search

mit-han-lab / data-efficient-gans

[NeurIPS 2020] Differentiable Augmentation for Data-Efficient GAN Training
https://arxiv.org/abs/2006.10738
BSD 2-Clause "Simplified" License
1.27k stars 175 forks source link

fmap-base and fmap-max meaning #69

Closed iumyx2612 closed 3 years ago

iumyx2612 commented 3 years ago

I see "fmap-base" in run_low_shot.py is "Number of feature maps" but in networks_stylegan2.py is "Overall multiplier for the number of feature maps". Same for "fmap-max", in run_low_shot.py is "Maximum number of feature maps", in networks_stylegan2.py is "Maximum number of feature maps in any layer". I assume fmap-max has the same meaning Can you explain "fmap-base" clearly?

zsyzzsoft commented 3 years ago

fmap-base is the overall multiplier for the number of feature maps. You can have a look at this line of code, which is the formula of determining the number of feature maps at each layer depending on fmap-base and fmap-max.

iumyx2612 commented 3 years ago

Oh, I got it, thank you