Deep Learning-Based Multi-Object Group Detection and Tracking in Video Streams

Uci, Erkan (2024) Deep Learning-Based Multi-Object Group Detection and Tracking in Video Streams. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (15MB) | Preview
Preview	PDF (Configuration Manual) Download (332kB) | Preview

Abstract

Accurately detecting and continuously tracking multiple object classes in video data datasets are critical challenges in computer vision, especially for autonomous vehicles and for applications such as video analytics. This project focuses on multi-object detection and tracking using the MOT17 dataset, leveraging the latest artificial intelligence and deep learning techniques to address these challenges.

Our methodology is a self-contained architecture using different advanced deep learning models to improve object detection and tracking accuracy. Convolutional Neural Networks (CNNs) are used to extract the correct features from video frames and to select objects of interest labeled. Long Short-Term Memory (LSTM) networks are included to preserve temporal dependencies and allow moving objects to be tracked seamlessly between frames. In addition, Faster R-CNN and R-CNN frameworks are integrated to improve object localization and classification through spatial and region-based recommendations and improved domain analysis.

To verify the reliability and robustness of our work, we conducted extensive experiments on various video datasets covering different environmental conditions and object densities. The results show that the combined use of CNNs, LSTMs, Faster R-CNNs and R-CNNs significantly improves the accuracy and reliability of multi-object detection and tracking.

By integrating computer vision and machine learning, this work provides important insights into their application in the real world, especially in autonomous vehicles and security systems. As mentioned before, the information gathered from different datasets and the proposed results using the methodology we have implemented will add significant value to the vision of high baccuracy and reliable multi-object detection and tracking.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Zahoor, Sheresh UNSPECIFIED
Uncontrolled Keywords:	Deep Learning; Computer Vision; Multi-Label Classification; Region Proposal Networks (RPN); Multi-Object Tracking; Object Detection
Subjects:	Q Science > QA Mathematics > Electronic computers. Computer science T Technology > T Technology (General) > Information Technology > Electronic computers. Computer science Q Science > QH Natural history > QH301 Biology > Methods of research. Technique. Experimental biology > Data processing. Bioinformatics > Artificial intelligence Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Artificial intelligence G Geography. Anthropology. Recreation > GV Recreation Leisure > Games and Amusements > Computer Games. Video Games. Q Science > QH Natural history > QH301 Biology > Methods of research. Technique. Experimental biology > Data processing. Bioinformatics > Artificial intelligence > Computer vision Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Artificial intelligence > Computer vision
Divisions:	School of Computing > Master of Science in Artificial Intelligence
Depositing User:	Ciara O'Brien
Date Deposited:	18 Jun 2025 14:52
Last Modified:	18 Jun 2025 14:52
URI:	https://norma.ncirl.ie/id/eprint/7927

Actions (login required)

View Item