Generative-Image-Inpainting-with-Contextual-Attention

The PyTorch reimplementation of the paper Generative Image Inpainting with Contextual Attention by Jiahui Yu, Zhe Lin, Jimei Yang, Xiaohui Shen, Xin Lu, and Thomas S. Huang, (2018) arXiv.

This implementation heavily relies on Contextual Attention Layer implementaion from https://github.com/daa233/generative-inpainting-pytorch.

Ideas for the experiments

1. Progressive growing mask
2. Multi-scale inference and network to merge the results
3. Sequential inference from outside inside

Also need to train on the Flickr dataset, to get more real-life results