Abstract:
Helmet detection plays a crucial role in ensuring the safety of motorcyclists. Helmet detection
system using the YOLOv5 architecture and integrate it into a Django web application is
proposed. The system aims to automatically detect the presence of helmets in images or real time video streams. The YOLOv5 model trained on a dataset consisting of helmet and non helmet images, utilizing transfer learning for improved performance. The trained model is
serialized and integrated into the Django application, allowing for seamless inference and
prediction of helmet detection. API endpoints are defined to receive image inputs and return
the detection results. The system includes data preprocessing steps to handle incoming images
and optimize them for the model's input requirements. The deployment and integration of the
model into the Django framework enable real-time helmet detection and provide a user-friendly
interface for users to interact with the system. Extensive testing and evaluation demonstrate the
effectiveness and accuracy of the helmet detection system. The developed solution has the
potential to contribute significantly to promoting helmet usage and enhancing road safety for
motorcyclists.
