The School of Computer Science is pleased to present…
Date: Thursday, April 10, 2025
Time: 11:00 am
Location: Essex Hall, Room 122
A steady increase in the number of vehicles on roads around the world has increased the need for Intelligent Traffic Systems (ITS). Vehicle detection, classification, and license plate recognition are essential for traffic analysis and ITS. License plate detectors are especially helpful to law agencies, as they assist in catching criminals by recognizing license plates. Systems employing Artificial Intelligence (AI) utilize image classification and object detection to monitor and analyze traffic on roads and highways. These systems are powered by state-of-the-art neural network architectures (e.g. the Convolutional Neural Network (CNN) for classification), which enable accurate detection and processing of real-time traffic data. Most vehicle monitoring systems, however, focus on only one aspect of vehicle tracking at a time. For example, these systems may focus on detecting specific attributes of a vehicle, such as the vehicle’s make, model, or color, often ignoring the other attributes. This manuscript introduces a novel approach to vehicle monitoring systems, which involves focusing on all such attributes simultaneously. We divide the Stanford Cars Dataset (SCD) into groups based on the car's make, model, type, year, and color. Subsequently, we train a separate CNN classifier on each group to learn the characteristics of each group. This allows us to study how each attribute independently affects the classification accuracy of the vehicle monitoring system.
Internal Reader: Dr. Dan Wu
External Reader: Dr. Mohamed Belalia
Advisor: Dr. Imran Ahmad
