NOVEL AI BASED APPROACH FOR HUMAN BODY TEMPERATURE EVALUATION USING INNER EYE CANTHUS LOCALIZATION FROM CAMERA FEED

Abstract

In this paper, we report development of an Artificial Intelligence (“AI”) based pipeline to localize inner canthus for human body temperature evaluation. Locating inner canthus in a camera feed remains a challenging endeavor. We have utilized numerous Facial points, nose points and eye landmarks to locate inner canthus and report four novel aspects of our work in evaluating human body temperature. Our first novelty lies in the process of locating ROI for correct observations. Our second novelty lies in solving the challenge of mapping RGB and Thermal images to get exact data points. Our third novelty lies in testing our algorithms/technology by developing automated testing pipelines on large datasets. Moreover, the fourth novel aspect of our work lies in our methodology to normalize the temperature from forehead, inner canthus & using our hardware configuration. We tested our algorithm with data comprising of different age groups, gender and geographical location and obtained 98.37% accuracy.