CNN based Face Recognition System for Patients with Down and William Syndrome

Abstract

Down syndrome, also known as trisomy genetic condition, is a genetic disorder that affects many people. Williams syndrome is a hereditary disorder that can affect anyone at birth. It marks medical and cognitive issues, such as cardiovascular illness, developmental delays, and learning impairments. This is accompanied by exceptional verbal abilities, a gregarious attitude, and a passion for music. Down syndrome and William Syndrome are both genetic illnesses. However, it can be distinguished from the arrangement of chromosome 21. Down syndrome and William syndrome can also be identified by recognizing faces, or facial characteristics, such as observing particular facial features. Therefore, this research develops Convolutional Neural Network (CNN) architectures to recognize Down syndrome and William syndrome using a facial recognition approach. A total of 480 facial photos were used in the study, with 390 images used for training data and 90 images used for testing data. The identification class is divided into three categories, Down syndrome, William syndrome, and normal. There are 160 photos in each patient class. This research presents two CNN architectures using a grayscale image of 256×256 pixels. The first CNN architecture comprises 12 layers, while the second comprises 15 layers. The average accuracy results with 12 layers were 91% by attempting to train and test six times. With 15 layers, the average accuracy value is 89%. In comparison, the first architecture has the highest accuracy value.