Hausa Visual Genome (HaVG)

Hausa Visual Genome (HaVG) 是由贝罗大学计算机科学与信息技术学院创建的一个多模态数据集，旨在支持英语到豪萨语的机器翻译研究。该数据集包含32,923张图片及其在豪萨语和英语中的描述，分为训练、开发、测试和挑战测试集。数据集的创建过程涉及自动翻译和人工后编辑，确保数据质量。HaVG适用于机器翻译、多模态研究及图像描述等多种自然语言处理任务，旨在解决豪萨语资源稀缺的问题。

Hausa Visual Genome (HaVG) is a multimodal dataset created by the School of Computer Science and Information Technology, Bayero University, aiming to support research on English-to-Hausa machine translation. It contains 32,923 images along with their descriptions in both Hausa and English, and is divided into training, development, test, and challenge test sets. The dataset was developed through automatic translation followed by human post-editing to ensure data quality. HaVG is applicable to various natural language processing tasks such as machine translation, multimodal research, and image captioning, and is intended to address the issue of scarce Hausa language resources.