Human Action Recognition in Office Environments

Abstract

This paper proposes an approach for classifying the actions of workers in office environments. The ultimate goal is to automatically calculate the working hours of workers and their other activities. Knowing what the workers are doing from each frame of the video during desk work makes them possible. In order to achieve this goal, we use You Only Look Once(YOLO) as an object detection method and Brute-Force Matcher as a prediction method. Using the proposed approach, videos are classified into six categories: "PC work", "calling", "writing", "stretching", "sleeping", and "others". In order to evaluate the proposed approach, we measure the accuracy by taking videos that assume desk work. The experimental results show that the proposed approach is more accurate than prediction using YOLO only.