Real-Time Emotion Detection and Mental Health Monitoring Using YOLOv11 with Explainable AI

Real-time emotion understanding in dynamic environments has brought increasing challenges of detecting the ongoing mental security problems that have arisen. Current systems hardly cope with their speed, accuracy, and interpretability for real-world applications in healthcare, education, and interactive technologies. This study proposes a powerful emotion and disorder detection in real time using YOLOv11. The model has been trained against large datasets like AffectNet labelled with seven key emotions anger, sadness, fear, surprise, neutral, disgust, and happiness. Preprocessing steps comprise resizing, normalization, and augmentation since it adds a quality assurance mechanism and increases the performance of the model. This system has been fine-tuned by optimized parameters and multi-component loss function to achieve high values of precision and recall. Explainable AI methods such as occlusion sensitivity maps also offer interpretability to those facial regions useful in determining emotions which provide clarity and trust.