Autism Spectrum Disorder (ASD), is a neurodevelopmental disorder. It is
characterized by impairment in social interaction, communications, restricted interests
and repetitive behaviors. It is important to detect this condition as soon as possible and
treat it. These differences helps to identify autism patient. ASD affects the physical
appearance of face. Autistic Children differ from children with regular development in
their patterns of facial features. With the aid of a web application that uses deep
learning to detect autism using their images and implements the Convolutional Neural
Network (CNN) algorithm to categorise children as autistic or typically developing, this
paper intends to assist families and psychiatrists in making an accurate diagnosis of
autism. We have used Tensorflow and Keras library in this process and lots of image
processing functions to smooth the model. This model can be implemented in hospitals
and clinics as GUI or software to detect autism disorder without any kind of medical
operations. retrained models used are VGG16 and VGG19. The facial photos were
acquired from a publicly accessible dataset on Kaggle that includes facial images of a
diverse range of kids, including autistic kids and kids without autism. The outcomes
were assessed using standard evaluation measures like accuracy, specificity, and
sensitivity. According to the classification findings' accuracy for the validation data,
VGG19 had the highest accuracy (96%), followed by VGG16 with an accuracy of
77.6%.
