Anticipation of Parking Vacancy During Peak/Non-peak Hours using Convolutional Neural Network – YOLOv3 in University Campus

Abstract

Searching for a publicly available parking space has become a nightmare to many drivers. With the constant development of global urbanization, human population has increased drastically in the past decades. Searching for a publicly available parking space in a highly populated area can be daunting and time consuming. No matter how much time is spent to find a vacant parking space, it always causes traffic congestion in the area. To alleviate these problems, it is of utmost importance to have a system that can detect and display the vacant parking spaces in real-time. This paper has conducted a study of anticipation of parking vacancy using convolutional neural network called YOLOv3 in a university campus. Image data is gathered from the video capture of the university’s campus open space parking lot. The YOLOv3 algorithm is used to train and predict whether the space is vacant or occupied. Results showed that YOLOv3 has been able to correctly predict the vacant space. The result of the rendering video will then be transformed into an image and is sent to the students via a Telegram group.