search

aitorzip / PyTorch-CycleGAN

A clean and readable Pytorch implementation of CycleGAN
https://arxiv.org/abs/1703.10593
GNU General Public License v3.0
1.19k stars 283 forks source link

question for the dataset structur #30

Open lucafei opened 3 years ago

lucafei commented 3 years ago

hi, i have a question for the dataset structure, because i'm new to the GAN, so i want to ask, the images from domain A and B must be one to one matched? another question is the number of images in domain A and domain B must be same?

BenBausch commented 3 years ago

As stated in the paper the input images do not have to be pairs. This is actually one of the main advantages of CycleGan.

SeucheAchat9115 commented 3 years ago

As stated in the paper the input images do not have to be pairs. This is actually one of the main advantages of CycleGan.

Hi, how is it possible to adapt the architecture for GTA sim10k to cityscape? GTA has 10k images and cityscape train has about 3k images, How could I adapt the architecture to fit for the unbalanced dataset? @aitorzip @BenBausch