LingoIITGN/triveni-raw

这个数据集结合了Vaani和Flickr30k两个来源的数据，是一个多语言和多模态的预训练语料库，同时还包括了一个用于微调的多语言、注释过的细调语料库。数据集支持多语言图像标题生成、图像-文本检索和视觉语境下的语音识别等任务。它包含了来自印度城市、文化、时尚、食物等多种类别的图像，并提供了三种语言的图像描述。

This dataset combines data from Vaani and Flickr30k sources, forming a multilingual and multimodal pretraining corpus, as well as a multilingual annotated fine-tuning corpus. It supports tasks such as multilingual image captioning, image-text retrieval, and speech-to-text grounding with visual context. The dataset includes images from various categories such as Indian cities, culture, fashion, food, etc., with captions provided in three languages.