kdd-miyaki/crag-mm-single-turn-public-v0.1.2

这是一个包含会话信息、图片以及交互对话的数据集，适用于图像理解和对话系统。数据集中的图片涵盖了多个领域，如时尚、书籍、科学、通用物体识别等。每个对话轮次记录了交互ID、领域、查询类别、动态性、查询内容和图片质量等信息。此外，还包括了答案和是否为自我中心视角的标识。数据集分为公开训练集和测试集。

This dataset contains session information, images, and interactive dialogues, suitable for image understanding and dialogue systems. The images in the dataset cover multiple domains such as fashion, books, science, general object recognition, etc. Each dialogue turn records information such as interaction ID, domain, query type, dynamism, query content, and image quality. Additionally, it includes answers and an indicator of whether it is egocentric perspective. The dataset is divided into public training and test sets.