Regresar

Detection of Motorcyclists Without a Safety Helmet Through YOLO: Support for Road Safety

Abstract:

In this research, the YOLOv4 algorithm based on deep learning was used, with the objective of detecting people who wear a safety helmet while driving a motorcycle, as well as those who do not use it while moving around the city, violating road safety regulations. In this context, the proposed methodology consisted of seven phases that go from the determination of the data source to the validation and deployment, in which the Labellmg tool and the Google Colab online platform are used due to their capacity and flexibility in the work environment. The model was developed using 287 images, of which 60% correspond to training images, 35% to validation and 5% to perform the tests. In addition, 30 additional photographic shots are available at different times of the day to determine the model's behaviour and precision. The results show that the trained model has a detection efficiency of 88.65% and that sing YOLOv5x could improve the detection quality by having a more significant number of layers.