arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2501.01243cs.CVcs.AIcs.CL
收藏

Face-Human-Bench:面向多模态助手的脸部和人体理解综合基准

Face-Human-Bench: A Comprehensive Benchmark of Face and Human Understanding for Multi-modal Assistants

Beijing University of Posts and Telecommunications(北京邮电大学);Beijing Normal University(北京师范大学)

Lixiong Qin, Shilong Ou, Miaoxuan Zhang, Jiangning Wei, Yuhang Zhang, Xiaoshuai Song, Yuchen Liu, Mei Wang, Weiran Xu

AI总结:

本文提出Face-Human-Bench基准,基于三层能力分类法构建1800个问题,评估25个多模态大语言模型的脸部和人体理解能力,并探讨能力相关性、位置影响及思维链提示效果。

摘要:

脸部和人体是社交互动中的关键元素,广泛存在于日常照片和视频中。因此,对脸部和人体的深入理解将使多模态助手能够提高响应质量并扩大应用范围。目前,多模态助手社区缺乏对脸部和人体理解能力的全面、科学评估。在本文中,我们首先提出一个包含三个能力层级的层次化能力分类法。然后,基于该分类法,我们从脸部和人体社区公开可用的数据集中收集图像和标注,并构建一个半自动数据流水线来为新基准生成问题。最终,得到的Face-Human-Bench包含一个开发集和一个测试集,每个集合有1800个问题,支持英语和中文。我们使用Face-Human-Bench对25个主流多模态大语言模型(MLLMs)进行了评估,重点关注能力之间的相关性、目标相对位置对性能的影响,以及思维链(CoT)提示对性能的影响。我们还探讨了MLLMs的哪些能力需要由专家模型来补充。数据集和评估代码已在https://face-human-bench.github.io公开提供。

英文摘要:

Faces and humans are crucial elements in social interaction and are widely included in everyday photos and videos. Therefore, a deep understanding of faces and humans will enable multi-modal assistants to achieve improved response quality and broadened application scope. Currently, the multi-modal assistant community lacks a comprehensive and scientific evaluation of face and human understanding abilities. In this paper, we first propose a hierarchical ability taxonomy that includes three levels of abilities. Then, based on this taxonomy, we collect images and annotations from publicly available datasets in the face and human community and build a semi-automatic data pipeline to produce problems for the new benchmark. Finally, the obtained Face-Human-Bench includes a development set and a test set, each with 1800 problems, supporting both English and Chinese. We conduct evaluations over 25 mainstream multi-modal large language models (MLLMs) with our Face-Human-Bench, focusing on the correlation between abilities, the impact of the relative position of targets on performance, and the impact of Chain of Thought (CoT) prompting on performance. We also explore which abilities of MLLMs need to be supplemented by specialist models. The dataset and evaluation code have been made publicly available at https://face-human-bench.github.io.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑