Adaptive deep learning based on FaceNet convolutional neural network for facial expression recognition

Facial recognition technology has become increasingly crucial in various applications, from personal identification, security, and human-care. Facial recognition has numerous practical applications, ranging from assessing mental health and well-being through facial expressions to evaluating customer satisfaction in service quality ratings. This study aims to develop a facial recognition model using a Convolutional Neural Network (CNN) with FaceNet architecture. The proposed method utilizes an advanced deep learning approach to generate high-quality facial embeddings, enhancing the model's ability to accurately identify and verify individuals. Our methodology includes training the CNN with FaceNet architecture, achieving an impressive average accuracy of 99.93%, with precision, recall, and F1-score all reaching 100%. The model demonstrated both high accuracy and efficiency, with an average training time of 13 minutes and 51 seconds. Future research should explore incorporating data augmentation, K-fold cross-validation, and additional transfer learning techniques to further enhance model performance and generalization. These advancements could lead to more resilient and flexible facial recognition systems capable of functioning effectively in diverse and challenging real-world conditions.