inpainting_cGAN

This repository is a Tensorflow implementation of "Conditional Attribute를 통한 Inpainting GAN 모델의 성능 개선"

Requirements

• tensorflow 1.9.0
• python 3.5.3
• numpy 1.14.2
• pillow 5.0.0
• scipy 0.19.0
• opencv 3.2.0
• pyamg 3.3.2
• opencv 4.1.0

Usage

Directory Hierarchy

inpainting_cGAN
├── Data/
│   ├── celebA/
│   ├── SVHN/
│   └── VUB/
└── src/
    ├── dataset.py
    ├── dcgan.py
    ├── download.py
    ├── image_edit.py
    ├── inpaint_main.py
    ├── inpaint_model.py
    ├── inpaint_solver.py
    ├── main.py
    ├── mask_generator.py
    ├── poissonblending.py
    ├── solver.py
    ├── tensorflow_utils.py
    └── utils.py

Download Dataset

You can use download.py to download datasets such as celebA and MNIST. You must put your dataset files under Data/ or you can manually set the directory in the dataset.py file.

Train GAN or cGAN

To train the model implemented in the dcgan.py file, run the next code.

> python main.py --is_train=true --dataset=celebA 

Train inpainting model

Attributions

• This project borrowed some code from semantic-image-inpainting and DCGAN-tensorflow.
• Special thanks to Sungkyunkwan University.