Implementation of Dlib Deep Learning Face Recognition Technology
Dujuan Zhang, Jie Li, Zhenfang Shan
Abstract
In order to overcome the problems of OpenCV in face detection, such as missing detection, false detection and poor recognition effect, a new method of Dlib face recognition based on ERT algorithm is proposed. This method can realize face recognition and feature calibration by Python, which calls a large number of trained face model interfaces, and it has good robustness for occlusion. By testing the process of face detection, feature point calibration, feature vector extraction and comparison in small deflections and positive faces images and videos, the experimental results show that the proposed method is superior to OpenCV method, it can effectively improve detection sensitivity, recognition precision and recognition effect. It can effectively solve the problem of poor real-time performance in dynamic image recognition.