TensorFlow implementation of "Context Encoders: Feature Learning by Inpainting" with CelebAMask-HQ Dataset.

The concept of 'Context Encoders' [1].

In this repository, 'Context Encoders' is trained with 'CelebA' Dataset [2].
The 'Context Encoders' consumes about 42 hours for training.

The 'Context Encoders' consumes 0.029 seconds for each sample in inference.

The results of 'Context Encoders' [1].

• Python 3.7.4
• Tensorflow 1.14.0
• Numpy 1.17.1
• Matplotlib 3.1.1
• Scikit Learn (sklearn) 0.21.3

[1] Deepak Pathak, et al. (2016). Context Encoders: Feature Learning by Inpainting. arXiv preprint arXiv:1604.07379.
[2] CelebA. http://mmlab.ie.cuhk.edu.hk/projects/CelebA.html
[3] CelebAMask-HQ Dataset. https://github.com/switchablenorms/CelebAMask-HQ