zjchust / ganimorph

Source code and information for the ECCV 2018 paper: Gokaslan et al., 'Improving Shape Deformation in Unsupervised Image-to-Image Translation'

GANimorph: Improving Shape Deformation in Unsupervised Image to Image Translation

This Git repository hosts the official code for 'Improved Shape Deformation in Unsupervised Image to Image Translation', by Aaron Gokaslan, Vivek Ramanujan, Daniel Ritchie, Kwang In Kim, and James Tompkin, to be published at ECCV 2018.

arXiv preprint

Coming soon

We are just setting up the repo, and we will fill in details about datasets and pre-build models soon. Thank you for your patience.

Bibtex

@inproceedings{Gokaslan2018,
  title={Improved Shape Deformation in Unsupervised Image to Image Translation},
  author={Aaron Gokaslan, Vivek Ramanujan, Daniel Ritchie, Kwang In Kim, and James Tompkin},
  booktitle={European Conference on Computer Vision},
  year={2018}
}

Dataset

The Microsoft Research Kaggle Cat vs. Dog Dataset serves as an excellent demo. Simply download the dataset and remove 128 examples from the holdout list.

Dependencies

Tensorflow > 1.2 (GPU strongly recomended) This repo uses the Tensorpack library at commit 01cab873feced688070d3ab113530f03c2525723 and all relevent dependencies.