CeleX-HAR
收藏资源简介:
CeleX-HAR数据集由安徽大学计算机科学与技术学院等机构的研究人员创建,是一个大规模的高清人类行为识别数据集。该数据集包含124,625个高分辨率事件流视频,覆盖150种日常行为类别。数据集的创建考虑了多视角、多种光照条件、动作速度和遮挡等因素。CeleX-HAR旨在支持高分辨率输入信号的网络结构设计,推动事件相机在人类行为识别领域的应用。
The CeleX-HAR dataset was developed by researchers from the School of Computer Science and Technology of Anhui University and other institutions. It is a large-scale high-definition human activity recognition dataset. This dataset contains 124,625 high-resolution event stream videos covering 150 categories of daily activities. The construction of the dataset takes into account multiple factors including multiple viewpoints, various lighting conditions, movement speeds and occlusions. CeleX-HAR aims to support the design of network architectures for high-resolution input signals and promote the application of event cameras in the field of human activity recognition.




