Vehicle Detection and Traffic Data Generation Using YOLOv8 in Metro Manila Highways / Imee Q. Compra, Simon Daniel M. Dela Cruz, Ronan M. Esguerra, Joshuel Ernest Q. Simbulan, Andrew James S. Tejerero.

Thesis (Undergraduate)

College of Science -- Bachelor of Science in Information Technology, Technological University of the Philippines, 2024.

Includes bibliographical references.

"The worsening traffic conditions in urban cities reflect neglect in proper preparation
and a lack of studies about traffic management. This problem can be pointed out as one of
the effects of having insufficient comprehensive data, specifically in Metro Manila. The
primary objective of this study is to develop a project that can replace the traditional ways
of collecting and generating traffic data with a more robust, automated, and scalable web
application. YOLOv8 – the main algorithm for vehicle detection – was implemented on a
diverse dataset of traffic images in highway environments. The model's accuracy, speed,
and robustness were assessed through precision, recall, and score metrics. Data generation
techniques are also employed to export traffic count values from either live or recorded
video inputs. The results demonstrate that the YOLOv8 achieved a high detection accuracy
with a mean average precision (mAP) of 86.76% and a real-time processing speed of 20-
30 frames per second (FPS), and successfully generating compact traffic data that can be
used for data-driven scientific studies. That concludes that this project is a potential
stepping stone towards understanding the underlying crisis involving road and
transportation systems as it takes advantage of modern technologies to resolve and provide
a faster and more flexible solution for the lack of extensive traffic volume data in cities." - Author's Abstract