Aplicación móvil para el reconocimiento de expresiones faciales en el autismo infantil.

This project developed a mobile application to recognize facial expressions in infants with autism, with the aim of enhancing educational and therapeutic interventions. The methodology involved an initial literature review to select deep learning models, followed by the collection of 2000 images categorized into four emotions (joy, anger, fear, and sadness) to train an image classification model based on MobileNetV2. The model was evaluated using metrics such as accuracy, model size, latency, and memory usage, ensuring its efficiency on mobile devices. During trials at the Special Education Unit "Fe y Alegría" in Santo Domingo, Ecuador, with children aged 4 to 9, the model achieved an accuracy of 74% on the training set and 70% on the validation set, demonstrating its ability to generalize to new data. The application was well-received by educators and therapists, who confirmed the accurate identification of emotions, enabling more effective and timely interventions. This tool holds significant potential to improve interaction and the educational process for children with Autism Spectrum Disorder (ASD), offering an accessible and user-friendly solution. It is recommended to expand the dataset to include a wider variety of facial expressions and contexts to further increase accuracy.

• Hoofdauteur: Aguirre Gómez , Jorge Alexánder (author)
• Formaat: bachelorThesis
• Taal: spa
• Gepubliceerd in: 2024
• Onderwerpen: Redes neuronales convolucionales; Aprendizaje profundo; Dataset de imágenes; Detección de emociones
• Online toegang: https://repositorio.uteq.edu.ec/handle/43000/9122