Unsupervised Color Enhancement

Tensorflow implementation for learning an image-to-image color enhancement using CycleGAN structure (unsupervised).

For image example:

It learns color affine transform function for each pixel in CIE L*a*b*. Network structure for transformation network looks:

This implementation is based on CycleGAN-tensorflow of xhujoy (https://github.com/xhujoy). This repository contains train and test codes for reproduce. Pretrained network model and dataset will be distributed soon.

---

Prerequisites

• tensorflow r1.0 or higher version
• numpy 1.11.0
• scipy 0.17.0
• pillow 3.3.0

Getting Started

Installation

• Install tensorflow from https://github.com/tensorflow/tensorflow
• Clone this repo:

git clone https://github.com/JunhoJeon/unsupervised-color-enhance
cd CycleGAN-tensorflow

Training and Test Details

To train a model,

CUDA_VISIBLE_DEVICES=0 python main.py --dataset_dir=/path/to/data/

Models are saved to ./checkpoints/ (can be changed by passing --checkpoint_dir=your_dir).

To test the model,

CUDA_VISIBLE_DEVICES=0 python main.py --dataset_dir=/path/to/data/ --phase=test --which_direction=AtoB/BtoA

Reference

• The tensorflow implementation of CyelcGAN (which this repository forked from), https://github.com/xhujoy/CycleGAN-tensorflow
• The torch implementation of CycleGAN, https://github.com/junyanz/CycleGAN
• The tensorflow implementation of pix2pix, https://github.com/yenchenlin/pix2pix-tensorflow