| Summary: | Face recognition is a technology that is used to identify and verify the identity of a person by their face. The process is carried out by matching human faces from an image or video against a database of faces. It consists of three stages: face detection, face capture (feature extraction) and face matching (classification). Face recognition is implemented in many applications nowadays. A few examples of face recognition applications are attendance systems and security systems. Due to the COVID-19 pandemic, the traditional and existing attendance systems are not suitable to be used because there is a risk of COVID-19 transmission through the attendance system. Moreover, the existing technology that is implemented in some applications is time-consuming and the common web authentication methods such as password-based and token-based authentication are insecure. In this project, the dataset was collected from 10 subjects. The collected images are cropped and resized to the required input size. In order to increase the dataset number, the resized images undergo image augmentation. During the classification, 5 types of deep learning-based models are trained and used to predict the subject's images. Then, the performance of the models (VGG16, VGG19, ResNet50, ResNet101 and ResNet152) is compared. Both VGG16 and VGG19 have a high validation accuracy of 100% compared to ResNet50 (96%), ResNet101 (94%) and ResNet152 (98%). However, VGG16 performed better than VGG19 as VGG16 had no misclassifications while VGG19 had one misclassification. © 2024 Author(s).
|
|---|