Real-time Zinara vehicle class identification system using Convolutional Neural Networks (CNNs)

Author

Matavire, Lisah.

Title

Real-time Zinara vehicle class identification system using Convolutional Neural Networks (CNNs)

Abstract

The research project aimed to develop a real-time deep learning system using Convolutional Neural Networks (CNNs) for classifying vehicles into predefined categories with high accuracy. Through rigorous testing and performance evaluation, the system achieved satisfactory results with an average accuracy of 93.8% across five vehicle classes. The error rates were notably low, indicating the robustness of the model. Both black box and white box testing revealed the system's ability to handle various scenarios and configurations. Additionally, precision, recall, and F1-score analysis demonstrated the system's effectiveness in accurately identifying vehicle classes. The accuracy for the five classes of vehicles i.e. Class 1,Class 2, Class 3, Class 4 and Class 5 were 96.4%,89.6%,94.7%,85.9% and 98.8% respectively While acknowledging the dataset size limitation, this study showcases the promising potential of CNN-based systems for real-time vehicle classification applications, contributing significantly to this evolving field. The study successfully realized this objective by developing a deep learning model leveraging CNN architecture. Through rigorous testing and evaluation, the system achieved high levels of accuracy, precision, recall, and F1-score, demonstrating its effectiveness in accurately identifying vehicle classes. However, the study acknowledges the limitation of the dataset size, which could impact the generalizability of the results. Nevertheless, the research provides a promising approach to automated vehicle classification, improving upon existing methods.

Date

June 2024

Publisher

BUSE

Keywords

Convolutional Neural Networks (CNNs),

Black box and white box testing,

Automated vehicle classification,

Supervisor

Mr. Kanyongo,