Human face feature point auxiliary labeling method

Abstract

The invention belongs to the technical field of video monitoring, and particularly relates to a human face feature point auxiliary labeling method, which comprises the following steps of: splitting human face feature points into main feature points and secondary feature points, pre-labeling the main feature points, performing auxiliary labeling on the secondary feature points, and manually judgingwhether the positions of the feature points need to be adjusted or not; wherein in the human face image, the main feature points are composed of points which are relatively fixed in position and canrepresent key information of the face; determining the secondary feature points by two adjacent main feature points and face edge information, wherein the positions of the secondary feature points canbe flexibly changed according to requirements. A main feature point pre-marking method and a secondary feature point auxiliary marking method are fully utilized to improve the marking efficiency; theapplicability of the labeling system is enhanced in a mode of combining the main feature points with different secondary feature point labeling schemes; labeling deviation caused by subjective factors is reduced through equidistant distribution constraints of the sub-feature points.

Description

technical field [0001] The invention belongs to the technical field of video monitoring, and in particular relates to a method for auxiliary labeling of human face feature points. Background technique [0002] Facial feature points contain rich face information, and are widely used in face recognition, pose estimation, expression recognition, 3D face reconstruction and other fields. There have been many studies in the field of face feature point detection. Among them, the method based on deep learning has brought qualitative improvement to the detection performance. However, the superior performance of deep learning depends on large-scale training data, and the general training set reaches more than 100,000 levels. . On the other hand, different application fields have different requirements for the number of facial feature points. For example, the number of feature points required for face alignment is small, and only 5 points are needed to rotate and zoom the image; Subt...

AI Technical Summary

Problems solved by technology

[0004] The purpose of the present invention is to propose a method for auxiliary labeling of facial feature points to solve the problems of low labeling efficiency, insufficient applicability and large influence of subjective factors on labeling accuracy in the prior art

Images