Multi-style face feature point detection method based on convolutional neural network

Abstract

The invention belongs to the technical field of face recognition, and particularly relates to a multi-style face feature point detection method based on a convolutional neural network. According to the invention, fine tuning training is carried out on a residual network (ResNet-152) by inputting multiple styles of face images, so that style discrimination features are obtained; style clustering isperformed on the original image by using K-means by using style discrimination features; a style-aggregated face image set is generated through a generative adversarial network; finally, the originalface image and the style aggregation face image are jointly used as input, and face feature point prediction is generated through a cascade strategy. According to the method, errors caused by multiple styles to face detection are reduced, the accuracy of face detection is improved, and the method can be applied to face recognition, head posture estimation, face reconstruction, 3D face reconstruction and the like.

Description

technical field [0001] The invention belongs to the technical field of face recognition, and in particular relates to a multi-style face feature point detection method based on a convolutional neural network. Background technique [0002] With the rapid development of science and technology and society, face recognition technology has been widely used. Such as finance, security construction, aerospace construction, student education and various entertainment and other fields. Recently, more and more people pay attention to the detection of facial landmarks, such as eye corners, eyebrows, and nose tips. It is also a prerequisite for computer vision applications. Facial landmark detection can be applied to a wide variety of tasks, for example, face recognition, head pose estimation, face reconstruction and 3D face reconstruction. At present, the latest progress in face landmark detection mainly focuses on learning discriminative features from rich deformations of facial sha...

AI Technical Summary

Problems solved by technology

[0003] However, the face feature point detection is easily affected by the multiple styles of the face and causes errors, such as grayscale images and color images, light and dark, strong contrast, dull contrast

At present, most of the face images used to train neural networks are from the wild, so the neglect of multiple styles of faces is becoming more and more serious.

Images