A Hybrid Architecture Based on Deep Learning for Object Recognition for Autonomous Driving

Autonomous Vehicle (AV) is the future of Auto industry, which integrates many high-end technologies. Autonomous Vehicle (AV) will enable the road transportation fully human-independent. In the context of driverless vehicles, smart driving assistance, and cutting-edge traffic assessment, object detection plays a crucial role. Real-time accurate object identification is crucial for traffic assessment and smart driving assistance. The system's primary duty is to provide the driver or controller with a precise understanding of the road or the area around the vehicle. Though visual based autonomous vehicles have demonstrated excellent prospects, there are few problems on how to find and interpret the difficult traffic conditions of the gathered statistics. Autonomous driving has been composed of a large number of different functions individually, such as vision based object detection. Further, a vision based object detection system is divided into dynamic and stationary objects on the road. In this study, a visual based system is intended to be designed for identification of different objects. The main contributions of this research are to detect multiple objects (both dynamic and stationary) that contribute to high and low risk of collision of vehicles on roads. Experimental results showed that our both trained models achieved higher precision compared to other state-of-the-art models.