Sistema De Detecci\u00f3n Temprana De Indicadores De Autismo Mediante Procesamiento De Im\u00e1genes y Roboflow

Autism spectrum disorder (ASD) is a neurodevelopmental condition characterized by social communication difficulties, repetitive behaviors, and atypical visual attention patterns. Early identification of ASD is essential to improving therapeutic intervention and long-term developmental outcomes. However, traditional evaluation methods rely heavily on clinical observations and specialist-administered questionnaires, often delaying diagnosis, particularly in areas with limited professional services. In this context, digital image processing and machine learning have emerged as promising tools to support diagnostic processes.This article proposes a computer vision\u2013based system that uses Roboflow for dataset management, annotation, and model training aimed at detecting visual indicators associated with early ASD traits, such as reduced eye contact, limited facial emotional response, and atypical body orientation. A standardized dataset containing ethically authorized images was created and processed through augmentation, normalization, and segmentation techniques to enhance model robustness. Training was performed using YOLOv8-based architectures optimized for real-time detection, complemented by a structured clinical questionnaire to increase classification accuracy Preliminary results show an accuracy above 85.4% in detecting relevant facial features, with stable performance across varied lighting conditions and mobile devices. These findings demonstrate the potential of combining computer vision with contextual information to support early ASD screening.