Face Recognition Systems: Comparation Point of View between CNN and LBPH Methods

Abstract

Face recognition systems have become increasingly prevalent in various applications, including security, biometric authentication, and digital identity verification. This article presents a comparative study on the implementation and performance of two face recognition methods: Convolutional Neural Network (CNN) and Local Binary Pattern Histogram (LBPH). The research utilized the Labeled Faces in the Wild (LFW) dataset, which comprises 19 classes of faces, with 760 images for training and 475 images for testing. The system was developed using the Python programming language, incorporating TensorFlow/Keras, OpenCV, and Visual Studio Code, along with a Graphical User Interface (GUI). The primary focus of this study was to implement both face recognition methods and analyze the selectivity of the system in distinguishing between known and unknown faces. Experimental results demonstrated that the CNN method offered superior classification stability and consistent face recognition, whereas the LBPH method provided faster training times and reduced computational complexity. Additionally, the results indicated that threshold settings significantly influenced the system’s ability to classify recognized and unknown faces. In conclusion, the study found that CNN is more suitable for applications requiring robust classification capabilities, while LBPH is better suited for lightweight face recognition systems that prioritize processing speed and efficiency.