DRIVEMATE: Empowering Safe Driving Through Real-Time Traffic Sign Detection and Speech Feedback on Mobile Devices Using YOLOv5 Algorithm and TensorFlow Lite

The aim of this research was to develop an application called DRIVEMATE, which serves as an assistive tool for mobile device users while driving. By leveraging the capabilities of computer vision and machine learning, this research aims to enhance driver awareness and promote safer driving practices. The ability to accurately detect and recognize traffic signs in real-time scenarios, coupled with speech-based feedback, provides drivers with timely and informative alerts, improving their decision-making and reducing the risk of accidents. The research employed two key features: text-to-speech functionality for speech output and live traffic sign detection. Agile scrum development methodology was employed by the developers to ensure a well-structured and timely development process. TensorFlow Lite was utilized to export machine learning models suitable for Android devices, while Google Colab facilitated the training of the machine learning model. The Flutter framework, in combination with Android Studio Code, was employed to construct the application. To evaluate its performance, the application underwent testing and evaluation based on the International Organization for Standardization (ISO) 25010:2011, focusing on functionality, usability, efficiency, and reliability. A survey was conducted involving 100 respondents, indicating that the application provided accurate results, its functions operated correctly, and it proved to be a usable and reliable tool for drivers. The integration of YOLOv5, a real-time object detection method, enhanced the application's ability to detect traffic signs efficiently. Real-time images were collected for training the machine learning model, ensuring accurate traffic sign detection and recognition with speech output.