Evaluation of Object Detection Models on Indian Road Scenarios for Pothole Detection

Road surface defects such as open manholes and potholes pose a significant risk to road users, often leading to accidents and vehicle damage. Timely detection of these hazards is essential for improving road safety and enabling intelligent transportation systems. This project, "Evaluation of Object Detection Models on Indian Road Scenarios for Pothole Detection," presents an embedded vision-based system for the real-time detection of open manholes in Indian road environments. The proposed system utilizes a Raspberry Pi as the central controller to process images, execute the trained object detection model, and control all connected hardware components. A web camera continuously captures the road ahead, and the deep learning model analyzes the captured frames to detect the presence of an open manhole. A DC motor represents the movement of a vehicle, while a motor driver controls the motor speed based on the detection results. When an open manhole is detected, the Raspberry Pi automatically slows down or stops the motor to prevent a potential collision. A 12 V battery supplies power to the complete system, ensuring portable and reliable operation. The proposed framework demonstrates the effectiveness of deep learning and edge computing for real-time road hazard detection, providing a cost-effective and intelligent solution for enhancing road safety and supporting autonomous and driver-assistance applications in Indian road scenarios