A Comparative Study on Face Recognition Using Deep Learning Approach

Biometric systems are utilized to examine and confirm an individual's identity for verification. There are various biometric methods such as fingerprint scans, voice recognition and iris scan are available. Face recognition is one of the significant methods that has been used in many kinds of applications for security and surveillance purposes nowadays. There are several methods available from the early days to recent times for face recognition. Deep learning is one of the most used techniques in different applications of computer vision. It is a technique that facilitates automatic feature learning and classifying of images. In this paper, a CNN-based framework has been proposed and evaluated with some of the transfer learning frameworks and with the Google Teachable Machine-created model using a newly created dataset of faces 1500 images. Among all the methods, MobileNetV2 and DenseNet169 transfer learning models obtained fine performance with an accuracy of 100% with almost no loss.