Based on Context Encoders: Feature Learning by Inpainting [1] and Semi-Supervised Learning with Context-Conditional Generative Adversarial Networks [2]
all environment settings are stored in requirements.txt
This project trains a model to predict a square patch lost of data inside an image. To execute the training the following code should be run.
python context.py
Example
python ccgan.py
Example
NOTE: For the generator a low resolution image is concatenated within the network
This project uses CIFAR10 data with only cats and dogs to reduce number of samples.
[1] Context Encoders: Feature Learning by Inpainting
[2] Semi-Supervised Learning with Context-Conditional Generative Adversarial Networks