Detecting Helmets of the Bike Riders Using Deep Learning Algorithms

Abstract

In our paper, we propose an approach for the automatic detection of helmets of bike riders in real-time. In our country, there are a large number of people who use bikes for their daily rides. Motorbikes are more popular than cars because they are less expensive to maintain, take up less parking space, and offer greater mobility and versatility in urban environments. Riding a bike is a lot of fun, but it can also be dangerous. Complete safety for bicycle riders is a primary goal of the proposed system. Despite the fact that helmet use is now legally required, many motorists continue to avoid donning them. Especially in developing nations, the mortality toll has been rising steadily over the past few years. To ensure the safety of the public, it is necessary to implement a system of helmet detection that can identify drivers who are not wearing protective headgear. For this mechanism, we use some real-time dataset about 3202. Here we collect wearing helmet 1911 and no helmet 1291 data and use algorithms like VGG16, Resnet50, MobileNet v.02, Inception V3, EfficientNet, and CNN. We got the highest accuracy from the EfficientNet about 98%. The implementation portion of the article also contains the methods used in each person's comparison statements. Model validation methods are also used in this investigation to create the best possible model for the given circumstances.