TencentARC/MM-RealSR

📖 MM-RealSR: Metric Learning based Interactive Modulation for Real-World Super-Resolution

Chong Mou, Yanze Wu, Xintao Wang, Chao Dong, Jian Zhang, Ying Shan
Tencent ARC Lab; Shenzhen Institutes of Advanced Technology, Chinese Academy of Sciences

If you want to use the paper model, please go to this Colab Demo for MM-RealSR .

🔧 Dependencies and Installation

• Python >= 3.7 (Recommend to use Anaconda or Miniconda)
• PyTorch >= 1.7

Installation

1. Clone repo

   git clone https://github.com/TencentARC/MM-RealSR.git
   cd MM-RealSR

2. Install dependent packages

   # Install basicsr - https://github.com/xinntao/BasicSR
   # We use BasicSR for both training and inference
   pip install basicsr
   # facexlib and gfpgan are for face enhancement
   pip install facexlib
   pip install gfpgan
   pip install -r requirements.txt
   python setup.py develop

Training (4 V100 GPUs)

python -m torch.distributed.launch --nproc_per_node=4 --master_port=4321 mmrealsr/train.py -opt options/MMRealSRNet_x4.yml --launcher pytorch --auto_resume

python -m torch.distributed.launch --nproc_per_node=4 --master_port=4321 mmrealsr/train.py -opt options/MMRealSRGAN_x4.yml --launcher pytorch --auto_resume

🏰 Model Zoo

Please download checkpoints from Google Drive or Github Release.

Testing

python inference_mmrealsr.py -opt (path to .yml file) --im_path (path to LR images) --model_path (path to checkpoint) --res_path (path to save SR images)

python Metric/LPIPS.py --folder_gt (path to HR images) --folder_restored (path to SR images)

python Metric/NIQE.py --folder_restored (path to SR images)

python Metric/DISTS.py --folder_gt (path to HR images) --folder_restored (path to SR images)

Results are in the results folder

Video Demo

Quick start can be found in Colab Demo for MM-RealSR .

python inference_mmrealsr_demo.py -opt (path to .yml file) --model_path (path to checkpoint) --im_path (path to LR image)

The visual results of noise and blur modulation:

BibTeX

@InProceedings{mou2022mmrealsr,
    author    = {Chong Mou and Yanze Wu and Xintao Wang and Chao Dong and Jian Zhang and Ying Shan},
    title     = { Metric Learning based Interactive Modulation for Real-World Super-Resolution},
    booktitle = {European Conference on Computer Vision (ECCV)},
    date      = {2022}
}

📧 Contact

If you have any question, please email [email protected].

🤗 Acknowledgement

Thanks to the following open-source projects:

• Real-ESRGAN.
• CResMD.
• CUGAN.