Vision Based Real Time Collision Detection Using MobileNetV2 and AI

Globally, one of the main causes of death is road accidents. In order to improve emergency response, there is a need of an effective and real-time accident detection system. To detect realtime accident, a deep learning-based approach proposed in this paper. It uses MobileNetV2 and camera modules placed in accident-prone locations. Live video streams processed by this system through deep learning models. Then, real-time alerts sent by it to the control rooms. Response time has been reduced also save lives with the help of this approach. Via heavy training, high strength and accuracy are realized. With a help of accident datasets, the system is also verified. To ensure that the MobileNetV2-based approach works well, this work is also compared with other machine learning approaches.