Comparative Analysis of Optimizers and Transfer Learning Methods for Headgear Detection Using Convolutional Neural Network

In untapped nations, cars are quite uncommon because motorcycles have always dominated the transportation industry. Over the past few years, there has been an increase in motorbike accidents. Motorcyclists who do not use fluorescent helmets because they don't think they offer enough protection are frequently engaged in traffic accidents. Today, it is against the law to operate a motorbike without a helmet, which increases accidents and fatalities in India. Every motorcyclist must wear a helmet when operating a motorcycle by law for the protection of the public. However, the majority of people continue to disprove this by driving without wearing helmets. In order to solve this problem, we provide a method in this work that uses real-time surveillance footage to automatically identify motorcycle riders without helmets and identify license plates. When the traffic police observe motorbike riders riding without helmets in a group at intersections, they also use CCTV footage to seize control of the drivers of those cars and fine the helmet-less riders. But only human initiative and dedication can make it happen. Because of its great accuracy, Convolutional neural networks are utilized for picture categorization and recognition. The Convolutional neural network uses a hierarchical model that builds a network in the shape of a funnel and eventually produces a fully connected layer where all the neurons are connected to one another and the output is processed. The 764 photos in the helmet detection dataset that is accessible on Kaggle have bounding box annotations in the PASCAL VOC format.