Course project for CS230, implemented IcGAN on MNIST and celebA
Data: Put MNIST and preprocessed celebA .npy files into data folder
Experiments: Modify the corresponding params.json file inside the experiments/base_model/ folder.
To train on MNIST dataset:
python main.py --OPER_FLAG <FLAG>
To train on celebA dataset:
python main.py --OPER_FLAG <FLAG> --celebA 1
Replace <FLAG> with
0 to train on IcGAN
1 to train encoder z of IcGAN
2 to train encoder y of IcGAN
3 to test IcGAN
Results: View the results in sample/celebA_gan or sampele/mnist_gan folder.
Tensorboard run tensorboardx:
tensorboard --logdir=<'path_to_log'>