Train your IQA models as simply as possible

An efficient and user-friendly template to help researchers and practitioners train IQA models with minimal setup.

🚀 🚀 🚀 News:

• To be updated...
• ✅ Aug. 11, 2025: We make this repository publicly available.
• ✅ Jul. 19, 2025: We create this repository.

TODO List 📝

• [] Provide a simple IQA benchmark on different IQA datasets.
• [] Collect awesome IQA models.
• [] Collect awesome datasets.
• [] Release inference templete.
• Release training templete.

Contents 📌

1. Introduction 👀
2. Reproducibility Baselines 🎉
3. Acknowledgement 💌

Introduction 👀

This repository provides a simple and efficient framework for training Image Quality Assessment (IQA) models. Our goal is to make it easy for researchers and practitioners to develop and evaluate your own IQA models with minimal setup.

Reproducibility Baselines 🎉

Experiments Settings ⚙️

TBU

Baselines 📑

表 1： The reproducibility results on different IQA benchmarks in terms of SROCC

Baseline	Synthetic IQA Dataset				Authentic IQA Dataset
	CSIQ	LIVE	TID2013	Kadid10K	BID	LIVEC	Koniq	SPAQ
HyperIQA
TeacherIQA
MANIQA
MobileViT-IQA
TOPIQ-NR
CLIPIQA

Acknowledgement 💌

We sincerely thank these following great public repositories:

• MoCo and PromptIQA : The code structure is partly based on their open repositories.
• IQA-PyTorch: This project is inspired by the great repository. And parts of the model architecture (CLIPIQA, TOPIQ_NR) are adapted from it.
• HyperIQA, MANIQA

Stars ⭐️

Citation 🖊️

If our work is useful to your research, we will be grateful for you to cite our repository:

@misc{simpleiqa,
  title={SimpleIQA: Train your IQA models as simply as possible.},
  author={Zewen Chen},
  year={2025},
  howpublished = "[Online]. Available: \url{https://github.com/chencn2020/SimpleIQA}"
}