Enhancing Rider Safety with Real-Time Object Detection on Smart Electric Bikes

This paper addresses the development, implementation and deployment of an advanced object detection system to enhance safety features on smart electric bikes. The project focused on improving real-time detection of traffic signs and potholes using a camera mounted on an electric bike. Deep learning models YOLOv8 and the Real-Time Detection Transformer (RT-DETR) were optimized for a Raspberry Pi 4. The system was evaluated for mean average precision, recall, and frames per second (FPS), with models optimized via ONNX conversion and quantization to meet computational constraints. Despite challenges like limited dataset size and hardware limitations, the YOLOv8 Nano model proved to be the most suitable, balancing accuracy and speed for real-time bike rider assistance. Future work includes expanding the dataset, using more robust hardware, and exploring model optimizations. This study establishes a foundation for future advancements in intelligent transportation systems for enhanced rider safety.