Tensorflow implementation of Learning to Discover Cross-Domain Relations with Generative Adversarial Networks.
Our NIPS paper ALICE improves DiscoGAN/CycleGAN/DualGAN, please see its code repo, and comparison on cartoon generation.
Alice4Alice: ALICE algorithms for painting the cartoon of Alice's Adventures in Wonderland
- Python 3.5
- Tensorflow 1.0
- Others
The demo is tested a toy dataset, with domain X as 5-component GMM, and domain Z as 2-component GMM. It can be easily extended to other GMM settings, and real dataset.
To train:
$ python DiscoGAN_main.py
- Official PyTorch implementation (https://github.com/SKTBrain/DiscoGAN)
- Taehoon Kim's PyTorch implementation (https://github.com/carpedm20/DiscoGAN-pytorch)