Diffusion-based Kernel Prior for Blind Super-Resolution (DIP-DKP)

Requirements

• Python 3.6, PyTorch >= 1.6
• Requirements: opencv-python, tqdm
• Platforms: Ubuntu 16.04, cuda-10.0 & cuDNN v-7.5

Quick Run

To run the code without preparing data, run this command:

cd DIPDKP
python main.py --SR --sf 4 --dataset Test

---

Data Preparation

To prepare testing data, please organize images as data/datasets/DIV2K/HR/0801.png, and run this command:

cd data
python prepare_dataset.py --model DIPDKP --sf 4 --dataset Set5

DIP-DKP

To test DIP-DKP (no training phase), run this command:

cd DIPDKP
python main.py --SR --sf 4 --dataset Set5

Results

Please refer to the report for results. DIP-DKP is randomly intialized, different runs may get slightly different results. The reported results are averages of 5 runs.

License & Acknowledgement

The codes are based on diffusion models, DIP and USRNet. Please also follow their licenses. Thanks for their great works.