Hybrid Model of Facial Attendance System based on Deep Learning

The authentication of faces has recently emerged as a promising topic in computer applications; thus, Facial recognition systems are growing rapidly. There is a need for face authentication systems in different sectors, such as education and commercial. The most trustworthy part of face authentication is biometrics, and the possibility of developing new authentication methods continues to spark the interest of academics. Digital images may suffer in quality and accuracy during face identification if there is a large variation when faces seem to be in different settings. This is because digital photos may be taken anywhere. The current research relies heavily on preprocessing images and deep learning with the LFW dataset. The present paper uses a hybrid deep unified model based on Inception and DenseNet. In this work, Images undergo two distinct processes during preprocessing: scaling and noise-filtering. During the deep learning-based classification process, the size is decreased to address the problem of excessive space and time consumption. The picture is subsequently treated using a noise-filtering method to improve image quality. To increase precision, noise is suppressed before training the hybrid model. Preprocessed facial images are then passed to Hybrid Model for training. To guarantee the trustworthiness of the face attendance system, the suggested model performance and accuracy have been compared to those of the standard models.