SESRG-InViSS: Image and video data set for human pose, action, activity and behaviour detection

摘要

This paper introduces the SESRG-InViSS image and video data set for human pose, action, activity and behaviour detection. The dataset consist of 804 images of human actor in various pose (standing front, standing right, standing left, squatting, sitting etc) and 210 videos of human actor/actors in various action (walking, associating and disassociating with an object, holding objects, dragging objects, entering and leaving the ambient), activities (taking out objects from the ambient, moving objects, swapping objects, surveying the ambient, inspecting objects) and behavior (normal and suspicious behaviour). This dataset was used extensively in the development of SESRG's InViSS (Intelligent Video Surveillance System), an intelligent integrated system for detection, identification and management of human activities from video surveillance recording and live video surveillance feed.