Face Recognition Model based on Deep Learning Method

Abstract

Facial recognition technology has made significant progress in recent decades and is now widely used across various fields. However, as application scenarios become more complex, traditional face recognition methods face limitations in handling intricate environments and varying facial expressions. This article aims to address these challenges by developing an efficient and accurate face recognition model using deep learning methods to enhance recognition accuracy and robustness. Key aspects include data collection and preprocessing, where an image dataset containing 90 different types of facial data was collected, preprocessed, and labeled, employing data partitioning and enhancement techniques. The model design involved creating (CNN) based on ResNet50, complemented by SVM for classification, followed by model training and optimization. To further enhance the model’s generalization and robustness, data augmentation and cross-validation methods were utilized. The model’s performance was evaluated using various metrics, including accuracy, recall, and F1 score, demonstrating its recognition capabilities. Additionally, visual diagrams depicting model loss, iterations, and accuracy were created to facilitate understanding and analysis of the training process. Through these research efforts, this paper achieved notable improvements in the accuracy and robustness of face recognition technology, providing a solid foundation for future research and applications.