Rapid Screening of Children With Autism Spectrum Disorders Through Face Image Classification

Autism spectrum disorders (ASD) impact the development of children’s language, motor, and expression abilities, causing great adverse effects on children’s growth. The incidence of autism screening is still quite poor, nevertheless, due to the traditional method’s time and financial requirements for child guardians. If symptoms of autism are detected early, children with autism usually return to normal development after effective medical intervention. Furthermore, the likelihood of accurately identifying children with autism grows if deep learning is used to recognize face images of autistic children. In this study, the dataset of autistic children’s faces in the Kaggle database [1] is selected to classify the typically developing children and autistic children through the face recognition model. On model selection, VGG19 [1], VGG16 [2], ResNet18 [3], ResNet101 [4], and DenseNet161 [5] are candidates. After training, among the five models, ResNet101 and DenseNet161 have better performance, and the recall rate of ResNet101 is higher in these two networks.