Shoe Detection Using SSD-MobileNet Architecture

Abstract

Falls are a global health issue that especially affects to the elderly. In our previous research, we used a millimeter wave radar to estimate the position of the feet for fall risk assessment of cane users. Radar sensors have a good accuracy, however, due to its low resolution, it is difficult to know if the radar is really tracking the position of the feet or of any other object. In this research, we present a shoe image detector using SSD-MobileNet architecture that could be used in combination with the radar to accurately track the position of the feet. The results show that the proposed detector could correctly recognize the position of the shoes in an image.