Published Paper Details:

The project presents a web-based application for the early prediction of Autism Spectrum Disorder (ASD) using both facial images and behavioral questionnaire data.The proposed system provides a fast, accessible screening tool by leveraging multimodal inputs. Facial images are analyzed using Convolutional Neural Networks (CNN) to detect autism-related facial patterns, while behavioral responses are processed using classical machine learning models such as Logistic Regression, Random Forest, SVM, and XGBoost. The application, built with Flask and Streamlit, allows users to upload images or enter questionnaire responses for real-time predictions, providing risk classification along with confidence scores. By integrating image-based and data-based predictions, the system enhances the accuracy and reliability of autism risk screening, supporting early awareness for parents, educators, and healthcare professionals. The model is trained on publicly available ASD datasets containing facial images and behavioral attributes, ensuring both diversity and robustness in prediction. Data preprocessing techniques such as normalization, augmentation, and feature scaling are applied to improve generalization and model performance. The platform's user-friendly interface ensures accessibility for non-technical users, making it suitable for preliminary home-based assessments. Moreover, the modular design of the system allows for easy updates and integration of additional data sources, including speech or text-based behavioral cues, in future versions.