Product recommendation system based on enhanced real-time face emotion recognition

Accurate facial expression recognition (FER) remains a challenging task in computer vision due to the complexity of human emotions and the variability of facial features under real-world conditions. The research proposes an enhanced real-time FER system utilizing an ensemble deep learning architecture that combines the feature extraction capabilities of ResNet34 with the classification strength of AlexNet. Trained and validated on the FER2013 benchmark dataset, the proposed ensemble model achieves a high classification accuracy of 80.83%, outperforming many existing individual CNN-based models. The system demonstrates strong generalization in detecting nuanced emotional states across diverse facial profiles and lighting conditions. Building upon this foundation, the model is integrated into a real-time product recommendation framework, where detected emotions are mapped to user-specific preferences to dynamically curate product suggestions at the point of login. This emotion-aware recommendation engine not only improves personalization but also enhances user interaction and engagement, offering significant implications for e-commerce, smart interfaces, and affective computing applications. Improves real-time comprehension of human facial emotions by combining two deep learning models. Uses a webcam to identify emotions and makes product recommendations based on the user’s present state of mind. Uses facial emotion recognition to provide a more intelligent and customized shopping experience.