Home // SMART 2023, The Twelfth International Conference on Smart Cities, Systems, Devices and Technologies // View article

Using Convolutional Neural Networks for Parking Sign Detection

Authors:
Parnia Haji Faraki
Hamid Reza Tohidypour
Yixiao Wang
Panos Nasiopoulos

Keywords: Autonomous Driving, Parking Sign Detection, Object Detection.

Abstract:
Automatic detection and classification of parking signs play an important role in autonomous and human-driven cars as it may lead to significant traffic reduction. Existing approaches mostly focus on traffic sign detection. Although there are a few studies in recent years that focus on parking sign detection, this field of study faces a lot of challenges such as the diversity of parking signs in different countries, the fact that the size of parking signs is usually smaller than that of normal traffic signs and the difficulty of understanding their meaning, a challenge that extends even to human drivers. This paper proposes a novel method for detecting and classifying parking signs using visual information. This study is conducted on a custom dataset of nearly 16000 images of parking signs in Vancouver, Canada. We base our approach on the YOLOv7X network, which is a powerful object detection algorithm, and obtained a mean Average Precision (mAP) of 95% on the test set, a notable result compared to the existing state-of-the-art object detection algorithm.

Pages: 1 to 5

Copyright: Copyright (c) IARIA, 2023

Publication date: June 26, 2023

Published in: conference

ISSN: 2308-3727

ISBN: 978-1-68558-071-1

Location: Nice, France

Dates: from June 26, 2023 to June 30, 2023