Facial Emotion Detection in Autistic Children: A Deep Learning Approach

Facial emotion detection in autistic children is a crucial area with significant implications for understanding social communication challenges and developing supportive technologies. It is an essential intervention for children with ASD (Autism Spectrum Disorder), as they have difficulty identifying or expressing their feelings. The system uses a pre-trained MobileNetV2 model to classify emotions from facial expressions in real time. The dataset, 'Autistic Children Emotions - Dr. Fatma M. Talaat', is used in this work, which includes six types of emotions. Data augmentation is applied to increase the diversity of the dataset. Meta-learning has been used during the model's training to ensure robustness and accuracy. The model also predicts emotions from live webcam input, displaying the results with confidence scores. The best-performing model is preserved using the ModelCheckpoint callback. It demonstrates the potential of deep learning for emotion recognition, with applications in healthcare, customer service, and human-computer interaction.