Optimization of Object Detection and Inference Time for Autonomous Driving

In order to solve the problem of object detection in autonomous driving environment, the Deep Learning-based Object Detector was separated into four areas: Stem Block, Backbone Network, Detector, and Extra Layer, and several deep learning optimization techniques were applied to each layer. The accuracy of the model and the Inference Time were conducted cost-effectively through the rich Recipient Filed compared to the computational complexity. This allows the autonomous in the environment, classification performance and accurate localization dnn based object detector the design. When comparing accuracy and speed in an autonomous driving environment with M2Det, a state of the art model of SSDs, the real-time object detector was 1.9 times faster, with a 1.4% difference in mAP.