Yor`ub´a Flickr Audio Captions Corpus (YFACC)

YFACC是一个针对尼日利亚低资源语言Yor`ub´a的语音-图像数据集，由斯坦伦博斯大学和布加勒斯特理工大学合作创建。该数据集包含6000条从Flickr8k数据集中选取的图像及其对应的Yor`ub´a语音描述，旨在通过视觉基础模型实现跨语言关键词定位。数据集的创建过程涉及图像选择、语音描述的翻译和录制，以及关键词的时间对齐。YFACC的应用领域主要集中在语言学研究，特别是对于无文字语言的记录和分析，以及开发适用于低资源语言的语音处理技术。

YFACC is a speech-image dataset targeting the low-resource Yorùbá language of Nigeria, co-developed by Stellenbosch University and Politehnica University of Bucharest. This dataset includes 6000 images selected from the Flickr8k dataset and their corresponding Yorùbá speech descriptions, aiming to enable cross-lingual keyword localization through vision-language foundation models. The dataset creation workflow covers image selection, translation and recording of speech descriptions, as well as temporal alignment of keywords. The application fields of YFACC primarily focus on linguistic research, especially the documentation and analysis of unwritten languages, as well as the development of speech processing technologies suitable for low-resource languages.