YOLO-based Tricycle Detection from Traffic Video
Amie Rosarie Cubeta Caballo, Chris Jordan G. Aliac · 2020
Traffic management is one of the challenging issues that need to be addressed by any urban area, one of which is Tagbilaran City, located at the Province of Bohol, Philippines. Thus, the need to employ measures such as traffic surveillance systems is imperative. In such kind of system, vehicle detection is a basic functionality and in this paper, a YOLO-based model is developed to detect a tricycle, which is a unique kind of public transportation. Training of the model is done on images of tricycles, extracted from actual traffic videos of selected intersections and the performance of the model is measured using the average precision. In this case, a 37.91% average precision is generated for tricycles. Increasing the number of annotated images of tricycles in the training dataset will produce a more precise detection model and including more of the other types of common vehicles will generate data that will support any traffic reduction measure.