Framework of our proposed method which consists of

• a dual-head feature extraction module
• a global color rendering module
• a context-aware retouching module
• a spatial attention based fusion

All modules are jointly learned from the annotated data in an end-to-end manner.

pip install -r requirements.txt

• MIT-Adobe FiveK Dataset
  • We use the setting of Image-Adaptive-3DLUT in our experiments, please refer to their page for details and data link.
• PPR10K

Prepare the dataset in the following format and you could use the provided FiveK Dataset class.

- <data_root>
    - input_train
    - input_test
    - target_train
    - target_test

Or you need to implement your own Class for your customed data format / directory arrangement.

The default settings of the most hyper-parameters are written in the parameter.py file. To get started as soon as possible (with the FiveK dataset), only the 'data_root' needs to be modified before training.

python train.py --data_root <path>

By default, the images, models, and logs generated during training are saved in save_root/dataset/name.

To evaluate your own trained model of a specific epoch, specify the epoch and keep the other parameters the same as training.

For example,

python evaluate.py --model *** --epoch 397

• Compared with other methods

• For video ./video