keras-IconGAN

paper: Adversarial Colorization Of Icons Based On Structure And Color Conditions : link

conditional dual GAN model for icon colorization implemented with Keras

official implementation (pytorch)

Examples

fake image	color image	shape image

Prerequisites

• Pillow
• tensorflow (2.5.0)
• tensorflow_addons (0.13.0)
• tensorboard (2.5.0)
• pickle
• opencv-python (for test.py)

Directory

├── preprocessed_data
│   ├── contour
│   ├── img
│   ├── labels
├── test_data
├── models
├── samples
├── logs
├── dataset.py
├── icongan.py
├── test.py
└── README.md

Dataset

icon, contour from https://github.com/jxcodetw/Adversarial-Colorization-Of-Icons-Based-On-Structure-And-Color-Conditions labels: used pickle

How to train

python icongan.py

How to test

python test.py --model_path=<model_path> --color_input=<color_img_path> --shape_input=<shape_img_path> --output_path=<output_path>