Performance Comparison of Several Deep Learning-Based Object Detection Algorithms Utilizing Thermal Images

Abstract

The recent advancement in deep learning approaches of machine learning and computer vision technology has paved the way for many advancements in object detection prediction models used in a variety of fields. Object detection algorithms are improving continuously. Many of the common Application Program Interfaces (APIs) or libraries can be used for this purpose. The most common techniques are implemented in Google TensorFlow object detection. Each Object detection technique has its advantages and disadvantages. Utilizing such models in thermal imaging is exceptionally necessary and required in systems such as automated driving and security system at night. A direct comparison between the most common state-of-the-art standard object detection methods helps in finding the best solution for thermal image detection/recognition systems. This paper discusses these algorithms and compares them in terms of accuracy and classification loss, we compare the performance of thermal image detection for the six most suitable object detection models that are supported by TensorFlow 2.0. Since we only have a small dataset (56) images, we enlarge it by cloning new images from the original ones with some variations in order to increase the number of images in the dataset. We perform a comparison among these models to select the most suitable algorithm for head detection utilizing thermal images, to be used to predict a human temperature entering a building or a campus.