HumorDB
收藏资源简介:
HumorDB是由伊利诺伊大学厄巴纳-香槟分校创建的一个专注于图形幽默理解的图像数据集。该数据集包含精心挑选的图像对,每对图像具有对比鲜明的幽默评分,强调触发幽默的微妙视觉线索,并减少潜在偏见。数据集大小为3545对图像,内容来源于在线图像库、漫画、社交媒体平台等,通过人工评估和机器学习技术创建。HumorDB旨在通过二元分类、范围回归和成对比较任务,捕捉幽默感知的主观性,并推动场景理解技术的发展。
HumorDB is an image dataset focused on graphical humor understanding, developed by the University of Illinois Urbana-Champaign. Comprising 3545 carefully curated image pairs, each pair features widely divergent humor ratings, highlighting subtle visual cues that trigger humor while mitigating potential biases. The dataset’s content is sourced from online image repositories, comics, social media platforms and other channels, and was constructed using human evaluation and machine learning techniques. HumorDB aims to capture the subjectivity of humor perception through binary classification, range regression, and pairwise comparison tasks, and advance the development of scene understanding technologies.




