README.md

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

Dataset	Dis Img	Ref Img	MOS	DMOS
LIVE	✓	✓		✓
A57	✓	✓		✓
LIVE_MD	✓	✓		✓
MDID2013	✓	✓		✓
CSIQ	✓	✓		✓
KADID-10k	✓	✓	✓(Note) ~~~~
TID2008	✓	✓	✓
TID2013	✓	✓	✓
CIDIQ_MOS100	✓	✓	✓
CIDIQ_MOS50	✓	✓	✓
MDID2016	✓	✓	✓
SDIVL	✓	✓	✓
MDIVL	✓	✓	✓
Toyama	✓	✓	✓
PDAP-HDDS	✓	✓	✓
VCLFER	✓	✓	✓
LIVE_Challenge	✓		✓
CID2013	✓		✓
KonIQ-10k	✓		✓
SPAQ	✓		✓
Waterloo_Exploration	✓	✓
KADIS-700k	✓ (code only)	✓

Basic Usage

0. Prerequisites

   pip install wget
   

1. General Python (please refer demo.py)

   from load_dataset import load_dataset
   dataset = load_dataset("LIVE")
   

2. PyTorch (please refer demo_pytorch.py)

   from load_dataset import load_dataset_pytorch
   dataset = load_dataset_pytorch("LIVE")
   

Advanced Usage

1. General Python (please refer demo.py)

   from load_dataset import load_dataset
   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)

   from load_dataset import load_dataset_pytorch
   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 more datasets: PaQ-2-PiQ, AVA, PIPAL, AADB, FLIVE, BIQ2021, IVC
• PyPI package
• HuggingFace dataset
• Provide more attributes
• Add TensorFlow support
• Add MATLAB support

Star History