Large Continuous Action Dataset (LCA)

Large Continuous Action Dataset (LCA) 是一个专为动作识别研究设计的大型数据集，由普渡大学电气与计算机工程学院创建。该数据集包含24个动作类别，视频内容为连续流媒体形式，包含多个动作实例，这些实例在时间和空间上经常重叠。数据集的视频拍摄背景有限，以减少背景对动作类别识别的提示作用。LCA数据集的创建过程涉及复杂的视频拍摄和人工标注，旨在通过提供多样化的动作实例和复杂的场景，推动动作识别技术的进步。该数据集特别适用于研究动作识别中的时间同步和场景重建问题。

The Large Continuous Action Dataset (LCA) is a large-scale dataset specifically designed for action recognition research, developed by the School of Electrical and Computer Engineering at Purdue University. This dataset includes 24 action categories, with videos presented in continuous streaming format and containing multiple action instances that often overlap both temporally and spatially. The videos in this dataset are captured with limited backgrounds to reduce distracting background cues for action category recognition. The development of the LCA dataset involves sophisticated video capture and manual annotation, with the goal of advancing action recognition technology by providing diverse action instances and complex scenarios. This dataset is particularly well-suited for investigating temporal synchronization and scene reconstruction challenges within the domain of action recognition.