NORMA eResearch @NCI Library

Traffic Sign Detection Using Deep Learning Algorithms

Khandalkar, Sanika Atul (2022) Traffic Sign Detection Using Deep Learning Algorithms. Masters thesis, Dublin, National College of Ireland.

[thumbnail of Master of Science]
PDF (Master of Science)
Download (767kB) | Preview
[thumbnail of Configuration manual]
PDF (Configuration manual)
Download (1MB) | Preview


Traffic sign detection, along with road architecture, is a vital component of the advanced driver assistance system. Previously, typical traffic sign recognition system algorithms were used, however, due to a more advanced and upgraded globe. The only traditional model available for detecting traffic signs was insufficient. At the same time, because of its limitations, diving deeper into the recognition or detection aspect was a little tough. Deep learning methods, on the other hand, were introduced for the implementation and detection of traffic signs. The dataset was obtained from Kaggle. The dataset includes over 50,000 traffic sign photos and 43 identified classes. A transfer learning-based model for traffic sign detection and implementation was presented in this work, which steadily minimizes the quantity of training data required. when compared to other machine learning or deep learning models, while also reducing computational expenses by using transfer learning models such as Resnet 50, VGG16, VGG19, and so on. In addition, the Convolution neural network has been proposed, in which layer-wise feature extraction was conducted using multiple convolutions and pooling processes, which were then compared and analyzed. Finally, the transfer learning-based model is retrained numerous times using statistical analysis and fine-tuning parameters at different learning rates. when CNN and Adam Optimiser were utilized, the results reveal that the CNN model can get up to 98.95 percent recognition rate in traffic sign identification. The classification and evaluation report comprises accuracy, F-1 score, and recall, as well as several additional aspects tailored for review and analysis. This research might aid in the identification of different forms of traffic infrastructure, such as roadway marking and roadside provisions. Python tools such as matplotlib, Keras, seaborn, pandas, and many others were used to run and deploy the code. For programming, Google collab and Jupyter notebook were also used.

Item Type: Thesis (Masters)
Uncontrolled Keywords: Deep learning; CNN; VGG16; VGG19; Resnet50; Traffic sign; multiclassification
Subjects: Q Science > QA Mathematics > Electronic computers. Computer science
T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science
T Technology > TL Motor vehicles. Aeronautics. Astronautics
Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Machine learning
Divisions: School of Computing > Master of Science in Data Analytics
Depositing User: Tamara Malone
Date Deposited: 21 Feb 2023 13:51
Last Modified: 02 Mar 2023 09:47

Actions (login required)

View Item View Item