Champion-Solution-for-CVPR-NTIRE-2024-Quality-Assessment-on-AIGC Quality Assessment for AI-Generated Content - Track 1: Image

Beijing University of Posts and Telecommunications.
Beijing Xiaomi Mobile Software Co., Ltd.

Introduction

The official repo of AIGC Image Quality Assessment via Image-Prompt Correspondence. (CVPRW2024, the first place in the image track of the NTIRE 2024 Quality Assessment for AI-Generated Content challenge).

Environment Installation

• torch 1.8+
• torchvision
• Python 3
• pip install ftfy regex tqdm
• pip install git+https://github.com/openai/CLIP.git

Data Preparation

Download the competition test dataset from the specified website and unzip it into the "./data/AIGCQA-30K-Image/test" directory.

Trained Datasets

Download AGIQA-1K, AGIQA-3K, AIGCIQA2023 and AIGCQA-30K-Image datasets and unzip them into the "./data" directory.

Training and Testing

After preparing the code environment and downloading the data, run the following codes to train and test model.

#AIGCQA-30K-Image
python train_aigcqa30k.py
#AGIQA-1K
python train_aigc_agiqa1k.py
#AGIQA-3K
python train_aigc_agiqa3k.py
#AIGCIQA2023
python train_aigc_aigciqa2023.py

For AIGCQA-30-Image dataset, run the following codes to get val and test output.

AIGC_DB_AIGCQA30K_VAL.py
AIGC_DB_AIGCQA30K_TEST.py

Citation

If you find our work useful in your research, please consider citing our paper:

@InProceedings{Peng_2024_CVPR,
    author    = {Peng, Fei and Fu, Huiyuan and Ming, Anlong and Wang, Chuanming and Ma, Huadong and He, Shuai and Dou, Zifei and Chen, Shu},
    title     = {AIGC Image Quality Assessment via Image-Prompt Correspondence},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops},
    month     = {June},
    year      = {2024},
    pages     = {6432-6441}
}

Our other works:

• "Thinking Image Color Aesthetics Assessment: Models, Datasets and Benchmarks.", [pdf] [code], ICCV 2023.
• "EAT: An Enhancer for Aesthetics-Oriented Transformers.", [pdf] [code] ACMMM 2023.
• "Rethinking Image Aesthetics Assessment: Models, Datasets and Benchmarks.", [pdf] [code] IJCAI 2022.