A Deep Learning Approach to Detect Unauthorized Person

Abstract

This analysis aims to identify Unauthorized Persons to ensure the protection of an organization. Essentially, this is a hybrid of two forms of detection. The first detects known and unknown faces, while the second detects ID cards. An image dataset should be connected with the model. Using the dataset images, we were able to identify faces and ID cards. The face recognition library and Histogram of Oriented Gradients (HOG) were used to recognize the faces, and the You Only Look Once (YOLO) V3 model was used to detect the ID card. From the test image our model will be able to recognize both the face and the object. Our model will show the person's name and draw a rectangular green shape around the face region if it recognizes faces. If the face does not match the encoded faces, the system will mark it as unknown, and if the face has covered by a hand or other object, the system will mark it as unknown and draw a red shape around the face area. If our model detects the Id card, it will make a rectangle shape in the detected region and color it green, just as it does with faces. If the model is unable to detect the identification card, it will draw a rectangle below the neck and paint it red.