Energy Aware Parking Lot Availability Detection Using YOLO on TX2

Abstract

Finding a parking space is a tedious and time-consuming task in a metropolitan city. Due to this problem, many researchers proposed an automatic parking lot occupancy detection system using a camera with a deep learning method to provide useful information in the smart city system. Since object detection for the parking lot is performed in real-time by utilizing CPU and GPUs while parking detection is working 24 hours a day and 365 days a year, therefore power saving is important to reduce the electricity cost. However, the energy-aware is not considered in most related works. In this paper, we proposed an energy-saving algorithm for parking lot availability detection using YOLO running on the TX2 machine. We experiment using small parking lot prototype and remote control cars. In the experiment, we compare our algorithm with the direct application of original YOLO for parking lot detection, the results show that it reduces power by 97 percent when there is no moving object in the parking lot area and 71 percent when there are moving objects in the parking lot area.