Mobile Perimeter Guard System based on Deep Convolutional Neural Networks

Abstract

In recent years, with the remarkable performance of CNN in the field of computer vision, various intelligent analysis tasks have been widely used in real world. At present, intelligent analysis tasks can be deployed both on GPU servers and on mobile terminals (e.g., smart phones, embedded boards). Compared with GPU server mode, mobile-terminal-based mode can not only greatly reduce the bandwidth, but also effectively reduce the calculation load of the server. More importantly, it can build a more stable intelligent analysis system. Based on the above considerations, a MobileNetSSD-based perimeter guard system is proposed and a series of performance evaluations are conducted in this paper. Through experimental evaluations, it is shown that the proposed perimeter guard system can be applied to the service of perimeter guard in a factory or a residential area.