Human Recognition by Machine Learning using Temperature Sensor

-Building an Original Machine Learning Model for Thermal Images-

Abstract

This paper describes a method for recognizing people using machine learning model trained from thermal images of people captured by a low-resolution temperature sensor. We used Darknet, a neural network framework with YOLOv7 implemented. As a result of conducting an experiment to recognize people using a learning model built using only temperature images, we succeeded in recognizing people with a detection rate of around 90% using a learning model trained with 10,000 iterations.