UDoc-GAN
UDoc-GAN copied to clipboard
Question about training detail
Hi, thank you for your paper and open-source code! The unpair method is really useful in practical scenarios.
I'm trying to train the model on my own dataset, and after reading the code, I have some questions.
- In dataset.py, there is an operation to
find a region that doesn't contain msk, What is the purpose of this operation and does it have a big impact on the final result? - How much do fake_A_buffer/fake_B_buffer affect the training?
I'd appreciate it if you could share some of your thoughts. Thanks for your time.
An explantion of buffer(ImagePool) purpose: https://github.com/junyanz/pytorch-CycleGAN-and-pix2pix/issues/75