Closed gaoyangyiqiao closed 6 years ago
The celeba dataset is less than 5 GB, which is not very large. I do not provide pre-trained model.
Do you mean this? Maybe things got changed, it is more than 20G and the dir is different from what you wrote.
The tar file of cropped and aligned images is less than 5gb, but the original one is much larger.
I have provided the preprocess.py
for preprocessing the original images. So you might have to download the original celeba dataset. Sorry!
Thanks a lot.
The data for training is large. Is there a pre-trained model?