Face feature point locating method

Abstract

The invention relates to a face feature point locating method comprising the following steps that a face is divided into five areas: the left eye and the left eyebrow, the right eye and the right eyebrow, the nose, the mouth and the external contour of the face, features points are arranged in the area of the whole face, and all the feature points are globally located by using a convolutional neural network; the feature points included in the four areas of the left eye and the left eyebrow, the right eye and the right eyebrow, the nose and the mouth of the whole face are locally located by using four convolutional neural networks; and the position coordinates obtained through local locating of the feature points included in the four areas of the left eye and the left eyebrow, the right eye and the right eyebrow, the nose and the mouth of the whole face are utilized to substitute the globally located position coordinates of the feature points of the corresponding areas so that the final position coordinates of all the feature points are obtained. The number of the used networks is low, and the number of the detected feature points is high with no requirement for additional attribute information of the face.

Description

technical field [0001] The invention belongs to the technical field of image recognition, and in particular relates to a face feature point positioning method. Background technique [0002] Face feature point positioning is based on face detection to further locate the eyes, eyebrows, nose, mouth and outline of the face. The main idea is to use the information near the feature points and the relationship between each feature point to locate . At present, there are mainly three methods for face feature point location: [0003] The first is a face five-point feature point detection algorithm based on a cascaded convolutional neural network. This algorithm uses 23 convolutional neural networks to form a cascaded system to locate the five points on the face. Its shortcomings are mainly Since a network is used to optimize the position of each feature point, if this algorithm is used for feature point positioning of 194 points, a large number of networks will be used, making the...

AI Technical Summary

Problems solved by technology

[0003] The first is a face five-point feature point detection algorithm based on a cascaded convolutional neural network. This algorithm uses 23 convolutional neural networks to form a cascaded system to locate the five points on the face. Its shortcomings are mainly Since a network is used to optimize the position of each feature point, if this algorithm is used for feature point positioning of 194 points, a large number of networks will be used, making the whole system very complicated

[0004] The second is a sixty-eight-point feature point detection algorithm based on a coarse-to-fine cascade network. This algorithm uses 16 convolutional neural networks to locate sixty-eight points on the face, according to the left eyebrow, left eye, The right eyebrow, right eye, nose and mouth divide the face area into six parts, and use the network for position positioning respectively. ...

Images