Motorcycle Helmet Detection and Usage Classification in the Philippines using YOLOv5 Algorithm

Abstract

Motorcycles are becoming the primary option for mobility worldwide, and the number of motorcycle riders has been exponentially increasing over the years. In the Philippines, it is reported that the increase in registered motorcycles was greater than the increase in total registered vehicles. However, motorcycles are notorious as one of the most dangerous and fatal modes of transportation. Hence, it is heavily enforced that motorcycle riders wear the proper motorcycle helmets that meet the safety standards of motorcycle riding. The study introduced a YOLOv5 algorithm-based motorcycle rider detection and helmet usage classification model. The study utilized two pieces of footage captured by the researchers in Makati City. The footage underwent frame segmentation and preprocessing before being loaded into the model for training. The results of the model showed a desirable performance in the detection and classification capabilities of the trained model. The optimal hyperparameter values were also found using the babysitting method for model validation. It is recommended that future studies ensure consistency in data samples to eliminate bias in any class of the model. The study also recommends using smaller increments for tuning the hyperparameter values to further investigate the effects of increasing and decreasing hyperparameter values and utilize separate models for the detection and classification tasks.