A Comparative Study of Convolutional and Transformer-Based Object Detectors Under Constrained Training Budget
MSc Thesis Proposal by:
Gauthami Ulhas Shirodkar
Date: 25 June 2026
Time: 12pm
Location: Essex Hall 122
Abstract:
Object detection is a core task in computer vision and is widely used in applications such as autonomous driving, medical imaging, surveillance, and intelligent monitoring. Recent advances in deep learning have led to the development of both convolutional neural network (CNN)-based and transformer-based object detectors. While these approaches have achieved remarkable performance, their effectiveness often varies across application domains, making model selection a challenging task for researchers and practitioners. This study presents a multi-domain benchmark of contemporary object detection architectures under constrained training budgets. Multiple state-of-the-art detectors are evaluated across diverse datasets representing both natural-image and medical-imaging domains. To ensure a fair comparison, all models are trained and evaluated under a consistent experimental framework using standardized training configurations. Performance is assessed using widely adopted object detection metrics, including mean Average Precision (mAP), precision, and recall. The benchmark provides a comprehensive comparative analysis of modern detection architectures and offers practical insights into their strengths and limitations across different domains, supporting future research and real-world deployment of object detection systems.
Keywords: Object Detection, Deep Learning, Computer Vision, Convolutional Neural Networks (CNNs), Transformer-Based Detection, Comparative Analysis.
Thesis Committee:
Internal Reader: Dr. Boubakeur Boufama
External Reader: Dr. Esam Abdel-Raheem
Advisor: Dr. Imran Ahmad
