Content Understanding and Reasoning Enhanced Small-Sample Blind Image Quality Assessment

This is the source code for CURE-IQA: Content Understanding and Reasoning Enhanced Small-Sample Blind Image Quality Assessment.

Dependencies and Installation

Pytorch: 2.11.0

CUDA: 13.0

Python: 3.11

For test:

1. Data preparation

To ensure high speed, save images and lables of each dataset with 'mat' files. Only need to run 'data_preparation_example.py' once for each dataset.

2. Load pre-trained weight for test

The models pre-trained on KonIQ-10k with 1%, 5%, 10%, 25%, 80% samples are released. The files in the folder of 'model' are obtained from open accessed source code of Long-CLIP. Please download the checkpoints from Long-CLIP, and put it into the folder of 'checkpoints'.

The pre-trained models can be downloaded from: Pre-trained models. Please download these files and put them in the same folder of code and then run 'test_koniq_rt'n'.py' to make intra/cross dataset test for models trained on n% samples.

For train:

The training code can be available at the 'Training' folder.

If you like this work, please cite:

{

author={Song, Tianshu and Cheng, Deqiang and Kou, Qiqi and Zhang, Sanyou and Huang, Yipo and Li, Leida},

journal={IEEE Transactions on Multimedia},

title={Content Understanding and Reasoning Enhanced Small-Sample Blind Image Quality Assessment},

year={2026, Early Access},

doi={10.1109/TMM.2026.3688392}

}

License

This repository is released under the Apache 2.0 license.