facebookarchive / NAM

Non-Adversarial Unsupervised Domain Mapping

NAM: Non-Adversarial Mapping

This repo contains PyTorch code replicating the main ideas presented in:

• NAM - Unsupervised Cross-Domain Image Mapping without Cycles or GANs
  Yedid Hoshen and Lior Wolf, ICLR 2018 Workshop
  ICLR Manuscript

• Non-Adversarial Unsupervised Domain Mapping
  Yedid Hoshen and Lior Wolf, ECCV 2018
  https://arxiv.org/abs/1806.00804

Examples

Edges2Bags

Top: DiscoGAN Middle: NAM: Bottom: Source

Edges2Shoes

Top: DiscoGAN Middle: NAM: Bottom: Source

Variation in Outputs:

Getting Started

1. Download Edges2Shoes data:

cd data
sh get_data.py
cd ..

2. Train DCGAN unconditional generative model for the A domain:

cd code
python train_gen.py

3. Use NAM to train a mapping from A to B:

python train_nam.py

4. Evaluate multiple image analogies:

python eval_variation.py $image_id 

Where $image_id is replaced with the ID of the image you wish to map.

Note: DCGAN training can diverge sometimes. Unconditional samples from each epoch are available in "code/unconditional_ims/". If DCGAN training diverged, simply re-run it.

License

This project is CC-BY-NC-licensed.