COIN
收藏资源简介:
COIN数据集是由清华大学自动化系和美图公司联合创建的大型视频分析数据集,专注于日常生活中的综合教学视频分析。该数据集包含11,827个视频,涵盖180个任务,分布在12个领域,如交通工具、电子设备等。数据集通过一个新开发的工具箱进行有效标注,每个视频都标有一系列步骤描述及其相应的时间边界。COIN数据集旨在通过提供丰富的语义分类和大规模视频样本,推动教学视频分析领域的深入研究,解决现有数据集在多样性和规模上的不足。
The COIN dataset is a large-scale video analysis dataset jointly created by the Department of Automation of Tsinghua University and Meitu Corporation, focusing on comprehensive instructional video analysis in daily life. It contains 11,827 videos covering 180 tasks across 12 domains such as transportation vehicles and electronic devices. The dataset is rigorously annotated via a newly developed toolbox, where each video is labeled with a series of step descriptions and their corresponding temporal boundaries. The COIN dataset aims to promote in-depth research in the field of instructional video analysis by providing rich semantic classifications and large-scale video samples, addressing the shortcomings of existing datasets in terms of diversity and scale.




