Enhanced MobileNetV1 for Early Autism Detection: A Transfer Learning Approach Using Facial Imagery

The complex neurodevelopmental disorder known as autism, or autism spectrum disorder (ASD), is typified by repetitive behaviors, verbal and nonverbal communication difficulties, and ongoing difficulties in social interaction. Since each person is affected differently and experiences symptoms that range from mild to severe, prompt and precise diagnosis is crucial for successful care. Automated solutions for ASD identification have emerged as a potential research subject thanks to developments in computer vision and machine learning.These solutions aim to bridge gaps in traditional diagnostic methods, offering faster and more accessible alternatives. They can also help minimize human error and variability in assessments, ensuring more consistent results.For people with ASD, early detection via automated methods can greatly enhance treatment results and quality of life. Image-based analysis provides a scalable, objective substitute for time-consuming, biassed traditional procedures like behavioral evaluations and clinical observations. This paper brings a model that can detect autism by image processing using MobilenetV1 as a feature extractor. In this study, the dataset is collected from the Kaggle website for testing the model, the dataset has a total of 2940 images of Autistic and Non-Autistic individuals. The study applies advanced preprocessing techniques and classifiers to enhance model performance. In terms of performance measurement, the model has been evaluated through accuracy, precision, recall, f1-score, and sensitivity. It achieves an accuracy of 88.33% with lower error rate 0.1167 which is comparably better than other studies.