MMIU
收藏资源简介:
MMIU数据集由苹果公司创建,专注于多模态助手中的视觉意图理解。该数据集包含12,000张图像和44,000个相关问题,旨在模拟用户向多模态助手提出的问题。数据集内容涵盖事实信息查询、本地商业搜索、食谱请求、导航指引等多个领域。创建过程中,通过标注者根据图像生成问题,并确定14种不同的意图类别。MMIU数据集的应用领域广泛,主要用于解决多模态助手在处理视觉信息时的意图识别问题,推动相关技术的发展。
The MMIU dataset was created by Apple Inc., focusing on visual intent understanding in multimodal assistants. It comprises 12,000 images and 44,000 associated questions, which are designed to simulate the queries users pose to multimodal assistants. The dataset covers multiple domains including factual information queries, local business searches, recipe requests, navigation guidance, and other scenarios. During its development, annotators generate questions based on image content and define 14 distinct intent categories. The MMIU dataset has broad application prospects, and is mainly used to solve the intent recognition problem when multimodal assistants process visual information, thereby promoting the development of related technologies.

- 1MMIU: Dataset for Visual Intent Understanding in Multimodal Assistants苹果公司 · 2021年



