PKU-AIGIQA-4K: A Perceptual Quality Assessment Database for Both Text-to-Image and Image-to-Image AI-Generated Images

作者：Jiquan Yuan, Jihe Li, Fanyi Yang, Xinyan Cao, Jinming Che, Jinlong Lin, Xixin Cao · 发表于：2025 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW) · 年份：2025 · DOI：10.1109/ICCVW69036.2025.00353 · 被引用次数：14 · 研究领域：Computer Science

ATGTQA, aimed at assessing the perceptual quality of AI-generated images (A/Gls), is essential for image quality screening and the evaluation of generative models. However, there are two issues in exsiting work: I) existing AIGIQA databases are limited to images generated by text-to-image (T2/) generative models, lacking exploration of the evaluation o_f images generated by image-to-image (12!) generative models; 2) existing image quality assessment (/QA) methods often fail to fully utilize the information provided by image prompts when simultaneously assessing T2/ and 121 AIG/s. To address these problems, we. first establish a large scale perceptual quality assessment database for both T21 and 121 AIG/s, named PKU-AIGJQA-4K. Second, we propose a novel partial-reference !QA (PR-/QA) method in this paper. Finally, leveraging the PKU-AIGIQA-4K database, we conduct extensive benchmark experiments using several pretrained models and the current !QA methods to explore the applicability of these methods for evaluating the quality o_f both the T21 and 121 A/Gls. The PKU-AIGIQA-4K database and codes are released on https://github.com/jiquan/23/AIGIQA4K.