This is an implementation of Globally and Locally Consistent Image Completion with PyTorch.  
日本語README（追加予定）

gif: input -> masked -> output -> blackout

• result1
  
• result2
  
• result3
  

• Python 3
• PyTorch
• NumPy
• OpenCV
• pyamg (optional, but required for post-processing)
• scipy (optional, but required for post-processing)

This code can run without GPU.

Download the trained model　to the current directory:

wget --continue -O completionnet_places2.t7 http://hi.cs.waseda.ac.jp/~iizuka/data/completionnet_places2.t7

Basic usage is:

python inpaint.py --input <input_image> --mask <mask_image>

The mask is a Black and White image. White is completed reagion. And, the mask should be the same size as the input image. If the mask is not specified, a mask with randomly generated holes will be used.
Mask example:

Other options:

• --gpu : Use GPU for the computation, however may not become faster. Defaults to False.
• --postproc : Able the post-processing(poisson blending). Dafaults to False. This poisson blending script is forked from this repository.

For example:

python inpaint.py --input example.png --mask example_mask.png

Finally, please push this repository Star bottom! Thank you.

• This is developed on a Linux machine running Ubuntu 16.04.
• According to the original paper, this pre-trained model was trained on Places2 dataset and thus performance is for natural outdoor images.
• This script referenced here and here.

This work is licensed under the Creative Commons:
  http://creativecommons.org/licenses/by-nc-sa/4.0/

Please cite the original paper in your publications:

@Article{IizukaSIGGRAPH2017,
  author = {Satoshi Iizuka and Edgar Simo-Serra and Hiroshi Ishikawa},
  title = {{Globally and Locally Consistent Image Completion}},
  journal = "ACM Transactions on Graphics (Proc. of SIGGRAPH)",
  year = 2017,
  volume = 36,
  number = 4,
  pages = 107:1--107:14,
  articleno = 107,
}