IQA-Dataset

IQA-Dataset copied to clipboard

A Unified Interface for IQA Datasets

This repository contains a unified interface for downloading and loading 20 popular Image Quality Assessment (IQA) datasets. We provide codes for both general Python and PyTorch.

Citation

This repository is part of our Bayesian IQA project where we present an overview of IQA methods from a Bayesian perspective. More detailed summaries of both IQA models and datasets can be found in this interactive webpage.

If you find our project useful, please cite our paper

@article{duanmu2021biqa,
        author = {Duanmu, Zhengfang and Liu, Wentao and Wang, Zhongling and Wang, Zhou},
        title = {Quantifying Visual Image Quality: A Bayesian View},
        journal = {Annual Review of Vision Science},
        volume = {7},
        number = {1},
        pages = {437-464},
        year = {2021}
        }

Supported Datasets

Basic Usage

1. General Python (please refer demo.py)

   dataset = load_dataset("LIVE")
   

2. PyTorch (please refer demo_pytorch.py)

   dataset = load_dataset_pytorch("LIVE")
   

Advanced Usage

1. General Python (please refer demo.py)

   dataset = load_dataset("LIVE", dataset_root="data", attributes=["dis_img_path", "dis_type", "ref_img_path", "score"], download=True)
   

2. PyTorch (please refer demo_pytorch.py)

   transform = transforms.Compose([transforms.RandomCrop(size=64), transforms.ToTensor()])
   dataset = load_dataset_pytorch("LIVE", dataset_root="data", attributes=["dis_img_path", "dis_type", "ref_img_path", "score"], download=True, transform=transform)
   

TODO

• [ ] Add support for the FLIVE dataset.
• [ ] Add support for the BIQ2021 dataset.
• [ ] Add TensorFlow support
• [ ] Add MATLAB support
• [ ] Provide more attributes if available.
• [ ] Add support for the IVC dataset.