search

rosinality / swapping-autoencoder-pytorch

Unofficial implementation of Swapping Autoencoder for Deep Image Manipulation (https://arxiv.org/abs/2007.00653) in PyTorch
Other
255 stars 47 forks source link

Are the hyper-parmaters same for different datasets? #10

Closed zrfreya closed 4 years ago

zrfreya commented 4 years ago

What's the learning rate, gradient penalty weight, training epochs for the Church, flickr waterfall/mountains, and other datasets?

rosinality commented 4 years ago

It seems like that authors have used same hyperparameters except image resolutions and batch sizes.

zrfreya commented 4 years ago

It seems like that authors have used same hyperparameters except image resolutions and batch sizes.

Could you please share the pretrained weights of the Church dataset? I have trained the network with the Church dataset. However, my results are worse than the results in your paper.

002672fff730e9216f9bd07d7e9be58849e04ef5_to_0052e05e97a799402b5da686943d7ca42d21ff22 009bc93861ffc9d11c906f8785cb95409fe92f79_to_00b02653aba8d23a0d4d8cd065bb1cad7e3ea0ba

rosinality commented 4 years ago

Actually I am not the author of the paper. Sorry for confusion.

zrfreya commented 4 years ago

Actually I am not the author of the paper. Sorry for confusion.

Oops. Sorry for my mistake. Thanks for sharing your great work.