Liang, Yu (Hugh); Wu, Dalei
College of Engineering and Computer Science
University of Tennessee at Chattanooga
Place of Publication
Single-camera multi-vehicle tracking is a preliminary step for traffic optimization. To perform vehicle tracking, a fine-tuned classification dataset was created from multiple sources. A ResNet classifier was trained, and its knowledge was utilized to build a localization dataset automatically without manual annotations. The dataset consists of videos generated from the MLK testbed. Three different object detection frameworks (SSD, Yolov3, and Yolov4) were evaluated with a pipeline that consists of the detector, the fine-tuned classifier and the tracker (DeepSort). Then the detector with the highest accuracy (Yolov3) was trained on the localization dataset. 90% of the ResNet classifier knowledge was distilled successfully in the final trained Yolov3 model. The tracker classification accuracy was further improved by proposing a correction methodology that considers both the camera's distance and limited data from the future frames.
I want to start with my endless love and respect for my father, whom I missed a few months before graduation. I also thank my mother and family for their continuous support throughout my life. I would like to express my deep gratitude to Dr. Mina Sartipi, my supervisor and mentor, for the last two years. I can not thank her enough. She allowed me to explore different and exciting areas with continuous help and unlimited support. She has always been my source of motivation and inspiration since I joined her group. Suppose I decided to go back to academia again and be a professor. In that case, this is the kind of professor I would like to be. I would also like to thank Dr. Thanh-Nam Doan, whom I work with closely during my masters. I learned a lot from him, and he was always coming up with ideas to improve the work. My sincere thanks to Dr. Yu Liang and Dr. Dalei Wu for being my panel members, for their valuable comments, and for instructing me in multiple courses. Moreover, my profound thanks to all my colleagues in CUIP, those who graduated and the current team. A special thank to Jose, Jin, Pete, Jermey, Reid, Austin, Kim, Katie, Yatri, Sree, Brennan, Misagh and the annotation team. I also want to thank the Sudanese community in Chattanooga to accommodate me as one of their family and make sure that I found myself a place of comfort, support, and belonging. And I would like to make a special mention to my colleague and friend Anas for making my transition to a new living environment a seamless process. Most of my current work wouldn't have been possible without working with Marwan, whose meeting has made a remarkable turning point in my thinking, life, and career. And finally, my deepest gratitude to my closest friends: Shaa, Sarah, Abdalla, and Nano.
M. S.; A thesis submitted to the faculty of the University of Tennessee at Chattanooga in partial fulfillment of the requirements of the degree of Master of Science.
Computer vision; Machine learning
ix, 54 leaves
Alharin, Alnour, "Deep learning-based fine-tuned multi-vehicle tracking with classification correction" (2021). Masters Theses and Doctoral Dissertations.
Available for download on Tuesday, May 31, 2022