SR-IQA (Slightly modified from IQA-Pytorch)
This evaluation code aims to facilitate fair comparisons between (real-world) image super-resolution papers.
- Modify the
path/to/resultsat Line 40 and Line 43 ininference.pyto the path of your SR-IQA folder.
git clone /~https://github.com/liyuantsao/SR-IQA.git
cd SR-IQA
pip install -r requirements.txt
python setup.py develop
# list all available metrics
pyiqa -ls
# test with arbitrary number of metrics
python inference.py --input [Your output image_path or dir] --ref [Ground truth image_path or dir] --save_file results/<EXP_NAME> -m psnr ssim lpips musiq maniqa clipiqa