YanJieWen / IGDFormer-light-up-dark

Pytorch implemention

IGDFormer-light-up-dark

Pytorch implemention

Dual attention Transformer based on illumination-guided for low-light image enhancement

Central south university

Overall Architecture

Getting Started

Environment

1. Clone this repo:

git clone https://github.com/YanJieWen/IGDFormer-light-up-dark.git
cd IGDFormer-light-up-dark-master/

Data Preparation

1. Download the dataset: [Retinexformer]
2. Create the new root and put datasets into parired_datasets.

Training

1. For single GPU train: Run the [train] and you can change the params follow your device.
2. For multi GPU and single PC: Run accelerate launch --multi_gpu train_engine.py
3. For more details, please check [HUGGING FACE]

Evaluation and demo

Make sure you have down load pretrained weights from Google, and put them into the save_weights root. Run [validation] or [demo]

Acknowledgement

Thank you for the surprise works that we follow：

1. BasicSR
2. Retinexformer
3. Restormer
4. Huggingface