Traffic Congestion Reduction through Real-time Object Detection: Analyzing the Effectiveness of different CNN models such as Mask RCNN, SSDNet and Yolo

Rastogi, Sanjay (2024) Traffic Congestion Reduction through Real-time Object Detection: Analyzing the Effectiveness of different CNN models such as Mask RCNN, SSDNet and Yolo. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (9MB) | Preview
Preview	PDF (Configuration Manual) Download (1MB) | Preview

Abstract

One of the major usage of this technology is in Vehicle detection and counting. Vehicle counting is the process of counting the number of different types of vehicles that have crossed a particular area/line or entered into a particular zone. The former is an extension of the latter, and there is no nice way of putting this, but these systems are practical as hell. For instance, the implementation of other classifiers to distinguish between diverse sorts of products like trucks, automobiles, and bicycles is an example. Other areas include traffic management and observation and-direction, highway and safety surveillance and directing, Urban planning and mapping, tracking of specific vehicles, automatic number plate recognition, real-time traffic information, tolls, congestion and related data, crowd/ pedestrian counting, facial recognition and alignment, and analysis to mention but a few. In our attempt to do this research, we are aim for a detailed evaluation of several objection models in traffic signals that are effective in counting the number of vehicles. Besides the accuracy of the vehicle calculation the priority is also on computation and space of the algorithms so that on can fit into the real time analysis. Performing a through analysis we realized that YOLOv10 together with YOLOv8 had a better performance compared to the rest of the models. Since this whole pipeline will act as a response system in integrating with the devices present in the traffic, comparing the overall performance in terms of the number of objects detected for YOLO against RetinaNet, SSDNet, and MaskRCNN, it was evident that YOLO fared better not only in detection but also the time taken to conduct the detection (higher FPS).

Item Type:	Thesis (Masters)
Supervisors:	Name Email Simiscuka, Anderson UNSPECIFIED
Uncontrolled Keywords:	Object Detection; Frame per Second; YOLO models; Real time analysis; Cloud Framework; Automated traffic management system
Subjects:	H Social Sciences > HE Transportation and Communications Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Cloud computing
Divisions:	School of Computing > Master of Science in Artificial Intelligence for Business
Depositing User:	Ciara O'Brien
Date Deposited:	02 Jul 2025 17:46
Last Modified:	02 Jul 2025 17:46
URI:	https://norma.ncirl.ie/id/eprint/7999

Actions (login required)

View Item