Cross-Domain-Image-Translation-Using-CycleGAN icon indicating copy to clipboard operation
Cross-Domain-Image-Translation-Using-CycleGAN copied to clipboard

Training Image Data

Open Saracenz opened this issue 5 years ago • 1 comments

Hi,

Do you know if the CelebA dataset is available already organised in the folders required?

trainA - images of male faces for training
trainB - images of female faces for training
testA - images of male faces for testing
testB - images of female faces for testing

I have downloaded the zip file of images but they are all in one big folder.

Saracenz avatar Jul 20 '20 08:07 Saracenz

Reading between the lines in the README, it looks like separating out the trainA, trainB, testA, and testB sets was left as an exercise to the user. You could programatically separate them using the annotation file which is a CSV with "male" being one of the annotations.

kindrowboat avatar Aug 11 '20 17:08 kindrowboat