search

yunjey / stargan

StarGAN - Official PyTorch Implementation (CVPR 2018)
MIT License
5.22k stars 968 forks source link

How to change dataset in the pre-trained StarGAN #55

Open Gigibulid opened 6 years ago

Gigibulid commented 6 years ago

How I can add my own dataset (even if that is one image) in the pre-trained model?? I changed the --dataset value from CelebA to the name of mine but it did not work.

SummerHuiZhang commented 6 years ago

I just change the --rafd_image_dir as mine 'Norland_Images' and CelebA need no change. It worked. "--dataset RaFD --image_size 256 --c_dim 5 --rafd_image_dir Norland_Images"

(However, each row of the samples are the same.)

Gigibulid commented 6 years ago

Thank you so much for your response @SummerHuiZhang !!!! However, I don't have the RaFD dataset... I only used the celebA. Is there any other way?

AndreasRef commented 6 years ago

I have the same question as @Gigibulid

SummerHuiZhang commented 6 years ago

@Gigibulid Firstly I think you need to add your photo into the celebA folder and then add the attribute of your photo into list_attri_celeba.txt. You know, the first row of list_attri_celeba.txt is number of your photos and second all attributes (like black_hair, laughing, sad). And the other rows are the same structure like "photo_name.jpg 1 1 -1 -1 1 -1 "(1 stands for that your photo has this attributr and -1 means not ). Hope it may help you.

ahkarami commented 6 years ago

Dear @SummerHuiZhang, Is there any code that one pass a single image into the network & then all possible states of it generated (via pre-trained model on CelebA Data Set)?