Person recognition from facial expressions with transfer deep learning methods

Recognition of people through images and videos has been an increasing subject of research in recent years. Face and object recognition technologies, which were developed with security concerns in the early periods, have gained more importance today in order to provide entertainment and economic benefits. As a result of the widespread use of mobile devices by almost every member of the society, the efforts of mobile device manufacturers to stay ahead of the competition in both security units and various software for digital images on mobile devices increase the interest in person recognition algorithms. In studies on face recognition algorithms, deep learning techniques have been widely used in image processing, which requires large data and complex calculations. In this study, Alexnet, Vgg16, Vgg19, Resnet50 and Resnet101 deep transfer methods were used in face recognition. The ORL dataset was used to test the methods. The highest success was observed with the Resnet101 model with %99,1. The lowest success was obtained with Vgg models.