ReInvade/LoRA-IR

[arXiv 2024] LoRA-IR: Taming Low-Rank Experts for Efficient All-in-One Image Restoration

LoRA-IR: Taming Low-Rank Experts for Efficient All-in-One Image Restoration

Yuang Ai^1,2  Huaibo Huang^1,2  Ran He^1,2

¹MAIS & NLPR, Institute of Automation, Chinese Academy of Sciences 
²School of Artificial Intelligence, University of Chinese Academy of Sciences 

arXiv 2024

⭐ If LoRA-IR is helpful to your projects, please help star this repo. Thanks! 🤗

🔥 News

• 2024.10.20: Release training&inference code, pre-trained models of Setting Ⅰ.
• 2024.10.20: This repo is created.

🏗️ Overall Framework

🔧 Dependencies and Installation

1. Clone this repo and navigate to LoRA-IR folder

   git clone https://github.com/shallowdream204/LoRA-IR.git
   cd LoRA-IR

2. Create Conda Environment and Install Package

   conda create -n lorair python=3.11 -y
   conda activate lorair
   conda install pytorch==2.3.1 torchvision==0.18.1 torchaudio==2.3.1 pytorch-cuda=12.1 -c pytorch -c nvidia
   pip3 install -r requirements.txt
   python3 setup.py develop --no_cuda_ext

⚡ Train & Inference

Training and Testing instructions for different settings are provided in their respective directories. Here is a summary table containing hyperlinks for easy navigation:

Setting	Training Instructions	Evaluation Instructions	Pre-trained Models
Setting Ⅰ	Link	Link	Download

🪪 License

The provided code and pre-trained weights are licensed under the Apache 2.0 license.

🤗 Acknowledgement

This code is based on NAFNet and BasicSR. Some code are brought from loralib, LLaVA and Restormer. We thank the authors for their awesome work.

📧 Contact

If you have any questions, please feel free to reach me out at shallowdream555@gmail.com.

📖 Citation

If you find our work useful for your research, please consider citing our paper:

@article{ai2024lora,
      title={LoRA-IR: Taming Low-Rank Experts for Efficient All-in-One Image Restoration},
      author={Ai, Yuang and Huang, Huaibo and He, Ran},
      journal={arXiv preprint arXiv:2410.15385},
      year={2024}
}