BIOLOGICALLY INSPIRED DEEP LEARNING SYSTEM APPLIED TO EGYPTION MULTI-STYLE LICENSE PLATE DETECTION

Abstract

This thesis presents a proposed biologically inspired deep learning system and performance on two tasks: first task is detection of Egyptian Car license plates. The second task is general object detection task in which we used the System to detect different objects in the Graz-02 object detection Dataset. the system consists of two main parts or stages: the first stage role is finding candidate object areas in the image. The second stage is mainly responsible for detecting precisely the objects among the candidate areas which are output from the previous stage. The system achieved detection percentage of 95% in detection of car plates task.