This repository was forked from https://github.com/Zeleni9/pytorch-wgan for the purpose of using WGAN in the context of training image based inversion.
This is the pytorch implementation of 3 different GAN models using same convolutional architecture.
- DCGAN (Deep convolutional GAN)
- WGAN-CP (Wasserstein GAN using weight clipping)
- WGAN-GP (Wasserstein GAN using gradient penalty)
The prominent packages are:
- numpy
- scikit-learn
- tensorflow 2.5.0
- pytorch 1.8.1
- torchvision 0.9.1
To install all the dependencies quickly and easily you should use pip
pip install -r requirements.txtRunning training of WGAN-GP model on 128x128 extractions from ti/zahner.png :
python main.py --model WGAN-GP-128 \
--is_train True \
--dataset ti_sampler \
--ti_file zahner.png \
--cuda True \
--batch_size 64 \
--dataroot ti
This training took around 4h 21 min on a single GPU.