AIGC图像质量评估指标研究
作者:
作者单位:

1.河南交通职业技术学院;2.河南师范大学 软件学院;3.黄河水利职业技术学院

中图分类号:

TP302.7

基金项目:

河南省高等教育教学改革研究与实践项目(2024SJGLX173,2019SJGLX690);河南省重点研发专项(231111210200,241111210300);中央引导地方科技发展专项(Z20221343001);黄河水利职业技术学院测绘地理信息职业教育研究课题(2021CHYB01)


Research on AIGC Image Quality Evaluation Indicators
Author:
Affiliation:

1.Henan College of Transportation;2.Henan Normal University;3.Yellow River Conservancy Technical Institute

  • 摘要
  • | |
  • 访问统计
  • |
  • 参考文献
  • | |
  • 引证文献
  • | |
  • 文章评论
    摘要:

    人工智能生成内容(AIGC)技术可为人类提供各种类型的信息生成服务,如何对AIGC进行准确的质量评估,是当前亟待解决的问题。本文主要针对大模型生成图像的质量及其评估指标开展深入研究。首先,从技术方面概述了当前评估AIGC的常见方法,如深度学习方法和计算机视觉方法等,介绍并分析了准确性、相关性、一致性、可解释性等指标在不同类型生成内容评估方面的表现。然后,为了展示评估指标的实际作用,以文心一言为例,对其生成的图像进行评估实验:使用直方图和噪点数量等量化指标对生成图像进行客观评估;使用整体协调性和美观性等视觉感官指标对生成图像进行主观评估。最后,综合对比客观评估和主观评估的结果,筛选出色偏、噪点数量、心理预期等AIGC产品质量评估的高可靠性指标。实验结果验证了综合使用主客观评估指标进行AIGC产品评估方法的有效性和可靠性。

    Abstract:

    Artificial Intelligence Generated Content (AIGC) technology can provide various types of information generation services for humans. The current urgent issue is how to accurately evaluate the quality of AIGC. This study conducts an in-depth research on the quality of images generated by large models and their evaluation metrics. Firstly, it summarizes the common methods for evaluating AIGC from a technical perspective, such as deep learning methods and computer vision methods. It focuses on introducing the metrics used in the evaluation methods, including accuracy, relevance, consistency, and interpretability, and analyzes their performance in evaluating different types of generated content. Then, to demonstrate the practical application of these evaluation metrics, this study conducts an evaluation experiment on the images generated by ERNIE Bot as an example. Objective evaluation of the generated images is carried out using quantitative metrics such as histograms and noise counts. Subjective evaluation is performed by assessing the overall coordination and aesthetic appeal of the images. Finally, by comparing the results of objective and subjective evaluations, this study screens out highly reliable metrics for evaluating the quality of AIGC images, such as color bias, noise count, and psychological expectations. This research not only provides a theoretical reference for evaluating the quality of AIGC, but also proves the effectiveness and reliability of using both objective and subjective evaluation metrics for AIGC product evaluation through experimental results. It has certain reference significance for obtaining feedback on model performance and optimizing performance.

    参考文献
    [1] 李白杨, 白云, 詹希旎等. 人工智能生成内容 (AIGC) 的技术特征与形态演进 [J]. 图书情报知识, 2023, 40(1): 66-74.LI Baiyang, BAI Yun, ZHAN Xini, et al. The technical characteristics and morphological evolution of artificial intelligence generated content (AIGC) [J]. Library and Information Knowledge, 2023, 40(1): 66-74.
    [2] WEN J, KANG J, XU M, et al. Freshness-aware Incentive Mechanism for Mobile AI-Generated Content (AIGC) Networks; proceedings of the 2023 IEEE/CIC International Conference on Communications in China (ICCC), F, 2023 [C]. IEEE.
    [3] ZHU Y, YANG F. ChatGPT/AIGC and Educational Innovation: Opportunities, Challenges, and the Future [J]. Journal of East China Normal University (Educational Sciences), 2023, 41(7): 1.
    [4] CAO Y, LI S, LIU Y, et al. A comprehensive survey of ai-generated content (aigc): A history of generative ai from gan to chatgpt [J]. arXiv preprint arXiv:230304226, 2023:
    [5] 徐振国, 王佳宁, 王悦等. 基于深度学习的自媒体平台在线课程质量评价研究 [J]. 电化教育研究, 2023, 44(08): 42-48.XU Zhenguo, WANG Jianing, WANG Yue, et al. Research on Quality Evaluation of Online Courses on Self media Platforms Based on Deep Learning [J]. E-education Research, 2023, 44(08): 42-48.
    [6] 陈向东, 褚乐阳, 王浩等. 教育数字化转型的技术预见: 基于 AIGC 的行动框架 [J]. 远程教育, 2023, 2:1672-0008.CHEN Xiangdong, CHU Leyang, WANG Hao, et al.. Technology foresight for digital transformation in education: an action framework based on AIGC [J]. distance learning, 2023, 2:1672-0008.
    [7] 王常圣. 人工智能驱动的数字图像艺术创作: 方法与案例分析 [J]. 智能科学与技术学报, 2023, 5(3): 406-414.WANG Changsheng. Artificial Intelligence Driven Digital Image Art Creation: Methods and Case Analysis [J]. Journal of Intelligent Science and Technology, 2023, 5(3): 406-414.
    [8] 李亚玲, 覃缘琪, 魏阙. 人工智能生成内容的潜在风险及治理对策 [J]. 智能科学与技术学报, 2023, 5(3): 415-423.LI Yaling, QIN Yuanqi, WEI Que. The potential risks and governance strategies of artificial intelligence generated content [J]. Journal of Intelligent Science and Technology, 2023, 5(3): 415-423.
    [9] 宋士杰, 赵宇翔, 朱庆华. 从 ELIZA 到 ChatGPT: 人智交互体验中的 AI 生成内容 (AIGC) 可信度评价 [J]. 情报资料工作, 2023, 44(4): 35-42.SONG Shijie, ZHAO Yuxiang, ZHU Qinghua. From ELIZA to ChatGPT: AI Generated Content (AIGC) Credibility Evaluation in Human Intelligence Interaction Experience [J] Intelligence data work
    [10] 吴柯烨, 孙建军, 谢紫悦. 基于专利文本挖掘的细粒度技术机会分析 [J]. 情报学报, 2023, 42(10): 1199-1212.WU Keye, SUN Jianjun, XIE Ziyue. Fine grained Technology Opportunity Analysis Based on Patent Text Mining [J]. Journal of Intelligence, 2023, 42(10): 1199-1212.
    [11] 张治. ChatGPT/生成式人工智能重塑教育的底层逻辑和可能路径 [J]. 华东师范大学学报 (教育科学版), 2023, 41(7): 131.ZHANG Zhi. ChatGPT/Generative Artificial Intelligence Reshapes the Bottom Logic and Possible Paths of Education [J]. Journal of East China Normal University (Education Science Edition), 2023, 41 (7): 131
    [12] 宋一飞, 张炜, 陈智能等. 数字说话人视频生成综述 [J]. 计算机辅助设计与图形学学报, 2023, 35(10): 1457-1468.SONG Yifei, ZHANG Wei, CHEN Zhineng, et al. Overview of Digital Speaker Video Generation [J]. Journal of Computer Aided Design and Graphics, 2023, 35(10): 1457-1468.
    [13] 宋萑, 林敏. ChatGPT/生成式人工智能时代下教师的工作变革: 机遇, 挑战与应对 [J]. 华东师范大学学报 (教育科学版), 2023, 41(7): 78.SONG Huan, LIN Min. The Work Transformation of Teachers in the Era of ChatGPT/Generative Artificial Intelligence: Opportunities, Challenges, and Responses [J]. Journal of East China Normal University (Education Science Edition) , 2023, 41(7): 78.
    [14] 于浩, 张文兰. ChatGPT 技术下教育面临的挑战和机遇 [J].中国医学教育技术, 2023, 37(3).YU Hao, ZHANG Wenlan. The challenges and opportunities faced by education under ChatGPT technology [J]. China Medical Education Technology, 2023, 37(3).
    [15] 严昊, 刘禹良, 金连文等. 类 ChatGPT 大模型发展, 应用和前景 [J]. 中国图象图形学报, 2023, 28(9): 2749-2762.YAN Hao, LIU Yuliang, JIN Lianwen, et al. Development, Applications, and Prospects of ChatGPT like Large Models [J]. Chinese Journal of Image and Graphics
    [16] WU F, HSIAO S-W, LU P. An AIGC-Empowered Methodology to Product Color Matching Design [J]. Displays, 2023: 102623.
    [17] LIU G, DU H, NIYATO D, et al. Semantic communications for artificial intelligence generated content (AIGC) toward effective content creation [J]. IEEE Network, 2024:
    [18] 吴虑, 杨磊. ChatGPT赋能学习何以可能 [J]. 电化教育研究, 2023, 44(12): 28-34.WU Lv, YANG Lei. How ChatGPT empowers learning is possible [J]. E-education Research, 2023, 44(12): 28-34.
    [19] Li C, Zhang C, Waghwase A, et al. Generative AI meets 3D: A Survey on Text-to-3D in AIGC Era [J]. arXiv preprint arXiv:230506131, 2023:
    [20] Zhang Z, Li C, Sun W, et al. A Perceptual Quality Assessment Exploration for AIGC Images [J]. arXiv preprint arXiv:230312618, 2023:
    [21] Wang T, Zhang Y, Qi S, et al. Security and privacy on generative data in aigc: A survey [J]. arXiv preprint arXiv:230909435, 2023:
    [22] 田阳, 王运武, 于燕娟等. ChatGPT 类人工智能推动教育数字化变革的 失范风险及应对策略 [J]. China Medical Education Technology, 2023, 37(3).TIAN Yang, WANG Yunwu, YU Yanjuan, et al. The risk of non-compliance and response strategies of ChatGPT type artificial intelligence in promoting digital transformation of education [J]. China Medical Education Technology, 2023, 37(3).
    [23] 王华树, 刘世界. 智慧翻译教育研究: 理念, 路径与趋势 [J]. 上海翻译, 2023, 170(3): 47.WANG Huashu, LIU Shijie. Research on Smart Translation Education: Concepts, Paths, and Trends [J]. Shanghai Translation, 2023, 170(3): 47.
    [24] 万小军. 智能文本生成: 进展与挑战 [J]. 大数据, 2023: 1.WAN Xiaojun. Intelligent Text Generation: Progress and Challenges [J]. Big data, 2023: 1
    [25] 祝智庭, 戴岭, 胡姣. 高意识生成式学习:AIGC技术赋能的学习范式创新 [J]. 电化教育研究, 2023, 44(06): 5-14.ZHU Zhiting, DAI Ling, HU Jiao. High Consciousness Generative Learning: Innovation in Learning Paradigm Empowered by AIGC Technology [J]. E-education Research, 2023, 44(06): 5-14.
    [26] 杨宗凯, 王俊, 吴砥等. ChatGPT/生成式人工智能对教育的影响探析及应对策略 [J]. 华东师范大学学报 (教育科学版), 2023, 41(7): 26.YANG Zongkai, WANG Jun, WU Di, et al. Analysis of the Impact of ChatGPT/Generative Artificial Intelligence on Education and Corresponding Strategies [J]. Journal of East China Normal University (Education Science Edition) , 2023, 41(7): 26.
    [27] Lu Z, Huang D, Bai L, et al. Seeing is not always believing: A Quantitative Study on Human Perception of AI-Generated Images [J]. arXiv preprint arXiv:230413023, 2023:
    [28] 周洪宇, 李宇阳. 生成式人工智能技术 ChatGPT 与教育治理现代化——兼论数字化时代的教育治理转型 [J]. 华东师范大学学报 (教育科学版), 2023, 41(7): 36.ZHOU Hongyv, LI Yuyang. Generative Artificial Intelligence Technology ChatGPT and Modernization of Educational Governance: A Discussion on the Transformation of Educational Governance in the Digital Era [J]. Journal of East China Normal University (Education Science Edition), 2023, 41(7): 36.
    相似文献
    引证文献
    网友评论
    网友评论
    分享到微博
    发 布
引用本文

邢润媚,常升龙,何宽,朱曙光,高琼,胡昊. AIGC图像质量评估指标研究[J].南京信息工程大学学报,,():

复制
分享
文章指标
  • 点击次数:282
  • 下载次数: 0
  • HTML阅读次数: 0
  • 引用次数: 0
历史
  • 收稿日期:2024-05-15
  • 最后修改日期:2024-07-11
  • 录用日期:2024-07-12

地址:江苏省南京市宁六路219号    邮编:210044

联系电话:025-58731025    E-mail:nxdxb@nuist.edu.cn

南京信息工程大学学报 ® 2025 版权所有  技术支持:北京勤云科技发展有限公司