BEOID

该数据集包含了一系列以自我为中心的活动视频，活动内容丰富多样，从烹饪到锻炼不等。此外，该数据集还采用了来自Ma等人的训练验证划分方法。其任务是进行时间动作检测。

This dataset contains a series of egocentric activity videos covering a diverse range of activities from cooking to exercise. Additionally, this dataset adopts the train-validation split method proposed by Ma et al. The task for this dataset is temporal action detection.