Optimized Algorithm for Face Recognition using Deepface and Multi-task Cascaded Convolutional Network (MTCNN)

Abstract

The combination of MTCNN and Deepface creates a robust and efficient facial recognition and recognition system to determine if a person is authorized. With 96.2% accuracy, MTCNN is known for its face detection and recognition accuracy. In addition to its excellent accuracy, MTCNN is also designed to operate in real-time, making it suitable for applications where instant face detection and recognition are critical. Additionally, MTCNN's offline capability increases its versatility, ensuring consistent performance even without an internet connection. On the other hand, Deepface acts as a light component in this combination. Although its accuracy is slightly lower at 95%, Deepface specializes in analyzing detailed facial features such as eyes, nose, and mouth. With this analysis, Deepface can extract facial features and provide additional information about the facial features of the recognized person as well as not recognized person. Despite its lightweight nature, Deepface's ability to analyze facial features adds depth to the facial recognition process and makes the system more comprehensive. Together, MTCNN and Deepface create a powerful synergy. The high accuracy and real-time capabilities of MTCNN combined with the detailed facial analysis of Deepface results in a system that offers both accuracy and efficiency. In addition, the combination ensures that the system works offline, making it reliable in various settings and conditions. Thus, this integrated approach provides an optimal balance between accuracy, lightweight design, and offline capability, making it well-suited for face recognition tasks where these factors are most important.