search

FunkyKoki / Look_At_Boundary_PyTorch

A PyTorch re-implement of CVPR 2018 LAB(Look At Bounday) --Champagne Jin
9 stars 1 forks source link

The pretrained discriminator? #10

Open YaN9-Y opened 5 years ago

YaN9-Y commented 5 years ago

Thanks for your excellent work! I want to continue training on your pretrained models of WLFW, but only generators are provided. Could you upload the pretrained discriminators together?

And are the models on WFLW trained with the default parameters of args.py? If I train it by myself with the default parameters, can I get the models of the same performance with the provided pretrained model at 900 epoch? Thanks!

FunkyKoki commented 5 years ago

It would be great if it helps you! The author didn't give the model structure of the discriminator, so I made one just like VGG16, you can try like this way, or you could try some better structure. @YaN9-Y

YaN9-Y commented 5 years ago

It would be great if it helps you! The author didn't give the model structure of the discriminator, so I made one just like VGG16, you can try like this way, or you could try some better structure. @YaN9-Y

Thanks for your reply! My expression might not be clear enough. I mean that I want to continue training on your pretrained models in https://pan.baidu.com/share/init?surl=tM3oJFUHmP4kJA7enXVLjA . But only weights of generators are provided, to continue training, the weights of discriminator is also needed. Since the training process is really slow, I wonder if you could upload the weights of pretrained discriminator altogether :D.
Besides, are the models you provide on https://pan.baidu.com/share/init?surl=tM3oJFUHmP4kJA7enXVLjA trained using default parameters on args.py?

Thanks!