Electronic Thesis and Dissertation Repository

Thesis Format

Monograph

Degree

Master of Engineering Science

Program

Electrical and Computer Engineering

Supervisor

Samarabandu, Jagath

Abstract

Traffic signs detection is becoming increasingly important as various approaches for automation using computer vision are becoming widely used in the industry. Typical applications include autonomous driving systems, mapping and cataloging traffic signs by municipalities. Convolutional neural networks (CNNs) have shown state of the art performances in classification tasks, and as a result, object detection algorithms based on CNNs have become popular in computer vision tasks. Two-stage detection algorithms like region proposal methods (R-CNN and Faster R-CNN) have better performance in terms of localization and recognition accuracy. However, these methods require high computational power for training and inference that make them difficult to apply in real-time applications. One-stage detection algorithms like Single Shot Multibox (SSD) and You Only Look Once (YOLO) are designed to be faster, but their accuracy is lower compared with the two-stage detector methods. In this project, a traffic sign detection algorithm is presented, which is inspired mainly by the SSD algorithm and its variants. The number of layers and the number of scales for object detection were modified to obtain the best balance in accuracy and speed detection. Experimental tests of this method over a traffic sign dataset give results of 93.75% mAP versus 89.35% mAP obtained using standard SSD+MobileNet, the speed of detection is 0.0124 s per image on a GPU.

Summary for Lay Audience

Traffic signs detection is becoming increasingly important as various approaches for automation using computer vision are becoming widely used in the industry. Typical applications include autonomous driving systems, mapping and cataloging traffic signs by municipalities. Deep learning algorithms have shown a state of the art performances in classification tasks. As a result, object detection algorithms based on deep learning have become popular in computer vision tasks. They can be divided into two main categories: Two-stage detection algorithms and one-stage detection algorithms. Two-stage detection algorithms have better performance in terms of localization and recognition accuracy compared with one-stage detection algorithms. However, one-stage detection algorithms are designed to be faster, which makes them suitable for real-time applications where the detection time is crucial. In this project, a traffic sign detection algorithm is presented, which is inspired mainly by state of the art one-stage detection algorithms. Modifications were made through experimentation to obtain the best balance in accuracy and detection speed.

Share

COinS