跨文化理解基准(CCUB)数据集
收藏资源简介:
跨文化理解基准(CCUB)数据集由卡内基梅隆大学机器人学院创建,旨在通过文化敏感的数据集改善文本到图像合成模型的文化代表性。数据集包含1095对来自8个不同国家的图像和文本描述,涵盖食物、艺术、宗教等多个文化类别。收集过程中,由文化专家根据其对文化的深刻理解挑选和描述图像。CCUB数据集的应用领域主要集中在提升AI生成图像的文化准确性和减少文化偏见,从而增强用户体验和文化尊重。
The Cross-Cultural Understanding Benchmark (CCUB) dataset was developed by the Robotics Institute of Carnegie Mellon University. It aims to improve the cultural representativeness of text-to-image synthesis models through culturally sensitive curated datasets. The dataset contains 1095 pairs of images and text descriptions from 8 distinct countries, covering multiple cultural categories including food, art, religion and others. During the data collection process, cultural experts selected and annotated the images based on their in-depth understanding of corresponding cultures. The primary application areas of the CCUB dataset focus on enhancing the cultural accuracy of AI-generated images and reducing cultural biases, so as to improve user experience and foster cultural respect.




