Multi-Camera Action Dataset (MCAD)

MCAD是由天津大学电子信息工程学院和新加坡国立大学交互与数字媒体研究所合作创建的多摄像头动作数据集，旨在评估监控环境下的开放视图分类问题。该数据集包含14,298个动作样本，涵盖18个动作类别，由20名参与者在5个独立摄像头下录制。数据集的创建过程涉及在不同时间和光照条件下重复录制每个动作，以模拟真实世界的多样性。MCAD的应用领域主要集中在动作识别技术，特别是在跨视图场景下的性能评估，以解决现有数据集在视图变化下的识别挑战。

MCAD is a multi-camera action dataset co-created by the School of Electronic and Information Engineering of Tianjin University and the Institute of Interactive and Digital Media at the National University of Singapore, which aims to evaluate open-view classification problems in surveillance environments. This dataset contains 14,298 action samples covering 18 action categories, recorded by 20 participants across 5 separate cameras. The dataset construction process involves repeating recordings of each action under different time and lighting conditions to simulate real-world diversity. The main application areas of MCAD focus on action recognition technologies, especially performance evaluation in cross-view scenarios, to address the recognition challenges under view variations faced by existing datasets.

搜集汇总

背景与挑战

背景概述

MCAD是一个多摄像头动作数据集，由天津大学和新加坡国立大学合作创建，包含14,298个动作样本，覆盖18个类别，由20名参与者在5个摄像头下录制，旨在模拟真实世界的多样性以评估监控环境下的开放视图分类问题。其应用主要集中在动作识别技术，特别是解决跨视图场景中的识别挑战，以弥补现有数据集在视图变化下的不足。