Social activity measurement with face detection using first-person video as a lifelog

This paper proposes a simple method of using video as a lifelog to measure the social activity of a camera wearer from a first-person perspective, aiming to quantify and visualize social activities that are performed unconsciously. The context of social activity is determined by the number and continuity of detected faces, whereas the amount of social activity is calculated by the number, size, and continuity. This taxonomy allows users to ascertain whether they tend to pass by people or spend time with them, and how many people there are. Our expectation is to enable users to change their behavior toward achieving social health by providing visual feedback. In this paper, we show an implementation of our social activity measurement and report on our feasibility study.