OVR
收藏资源简介:
OVR数据集由谷歌Deepmind开发,是一个用于视频中时间重复计数的大型开放词汇数据集。该数据集包含72,552个视频,每个视频都有重复次数、开始和结束时间的注释,以及自由形式的描述。数据来源于Kinetics和Ego4D,涵盖了广泛的行动和活动。数据集的创建过程包括从现有数据集中筛选视频片段,并进行大规模的众包注释工作。OVR数据集的应用领域广泛,旨在解决视频中重复行为的识别和计数问题,特别是在需要详细理解和分析重复行为的场景中。
The OVR dataset, developed by Google DeepMind, is a large-scale open-vocabulary dataset for temporal repetition counting in videos. It contains 72,552 videos, each annotated with repetition counts, start and end timestamps, as well as free-form descriptions. The dataset is sourced from Kinetics and Ego4D, covering a wide range of actions and activities. The creation of the OVR dataset involves filtering video clips from existing datasets and conducting large-scale crowdsourced annotation work. The OVR dataset has a wide range of application scenarios, aiming to address the problem of recognition and counting of repetitive behaviors in videos, especially in scenarios that require detailed understanding and analysis of such behaviors.




