Please use this identifier to cite or link to this item: https://hdl.handle.net/10356/72584
Title: Traffic sign classification based on binarized convolutional neural network
Authors: Lu, Chenyue
Keywords: DRNTU::Engineering::Electrical and electronic engineering
Issue Date: 2017
Abstract: Traffic sign classification in the traffic context is a crucial task for Intelligent Transportation Systems (ITS) and is becoming even more so under the current increasing pressures of transportation efficiency and road safety. Convolutional Neural Network (CNN) can be used in images classification; however, it usually requires the high-performance Graphics Processing Unit (GPU) and the large internal storage. Furthermore, the speed of CNN is difficult to meet the requirements of ITS. The Binarized Convolutional Neural Network (BCNN) is a pure binary system, in which the weights and activations are binarized. As a result, the efficiency of storage and recognition of ITS can be significantly improved through the use of the BCNN. Thus, the BCNN is adopting for traffic sign classification in this project. This project provides three main contributions. First, a critical review of the current state of obtaining the capability for traffic sign classification. This critical review presents the state-of-the-art research advancements and technologies need to be involved in this project. Second, a novel BCNN platform was developed. Both binarization function and binarized convolution function have been implemented in this toolbox. Third, the verification of the BCNN demo via Belgium Traffic Sign Classification Benchmark (Belgium TSC) and German Traffic Sign Recognition Benchmark (GTSRB) was conducted and presented. The efficiency and accuracy of the BCNN demo was discussed. Keywords: Traffic sign classification; Convolutional Neural Network; Binarized Convolutional Neural Network; BCNN toolbox development
URI: http://hdl.handle.net/10356/72584
Fulltext Permission: restricted
Fulltext Availability: With Fulltext
Appears in Collections:EEE Theses

Files in This Item:
File Description SizeFormat 
LuChenyue_2017.pdf
  Restricted Access
2.81 MBAdobe PDFView/Open

Page view(s)

137
Updated on Jun 23, 2021

Download(s)

14
Updated on Jun 23, 2021

Google ScholarTM

Check

Items in DR-NTU are protected by copyright, with all rights reserved, unless otherwise indicated.