sionic-ai/stvqa

该数据集包含图像和与其相关的对话信息。每个样本都有一个唯一的标识符(id)，一张图像(image)，以及一个或多个对话(conversations)。每个对话包括对话内容(content)和角色(role)。此外，每个样本还包含语言(language)和来源(source)的元数据(metadata)。数据集分为训练集(train)，共有18920个样本，大小为1019676709.56字节。

The dataset includes images and associated conversation information. Each sample has a unique identifier (id), an image (image), and one or more conversations (conversations). Each conversation includes the content (content) and role (role). Additionally, each sample contains metadata (metadata) including language (language) and source (source). The dataset is split into a training set (train) with a total of 18,920 samples, sized at 1,019,676,709.56 bytes.