Coarse-to-fine human face key point detection method

Abstract

The invention provides a coarse-to-fine human face key point detection method, which comprises the following steps of: 1, setting a set of human face key points, namely representing the whole face byusing 132 key points; 2, collecting a batch of unlabeled human face pictures as samples; 3, constructing a network, namely constructing a full-convolution neural network to carry out regression key point thermography; 4, training an overall rough model, a local accurate eyebrow model, a local accurate eye model, a local accurate nose model and a local accurate mouth model by using a neural networkstructure; and 5, inputting a picture, obtaining rough 132 key points through the overall rough model, obtaining accurate local five-sense-organ key points, and combining eyebrow, eye, nose and mouthkey points detected in the four local accurate models with contour key points detected by the overall rough model to obtain the accurate human face 132 key points. According to the invention, the precision of key point detection is improved, and the optimization effect on local five sense organs is obvious.

Description

technical field [0001] The invention relates to the technical field of face recognition, in particular to a coarse-to-fine face key point detection method. Background technique [0002] Face key point detection refers to inputting a two-dimensional face photo and locating the key point coordinates with specific semantics on the face in the photo. Common face key points include: face external contour key points, eyebrow contour key points, eye contour key points, nose contour key points, mouth contour key points, etc. figure 1 Examples of several common face key points are given. [0003] figure 1 : (a) The AR face dataset was released in 1998 and contains 22 key points of faces. (b) The XM2VTS face data set was released in 1999 and contains 68 face key points (c) the FRGC-V2 face data set was released in 2005 and contains 5 face key points. (d) The AFLW face dataset was released in 2011 and contains 21 face key points. (e) The AFW face dataset was released in 2013 and c...

AI Technical Summary

Problems solved by technology

Different data sets often use different keypoint specifications, which are difficult to use at the same time

[0009] 2. The number of face key points in most data sets is insufficient

For example, in the 68 key points of the face defined in the IBUG dataset, the key points of the eyebrow part only surround the upper edge of the eyebrow, while there are no key points on the lower edge; the nose part is also a very important face feature, which also lacks key points to Represent this feature; while the key points of the eyes and the outline of the face, which are important features of the face, are slightly sparse, and the ability to express the features of the face is limited

[0010] 3. Most of the faces in the data set are of European and American race, which is inconsistent with the actual application scenario of this patent, which is also a key factor affecting the accuracy of the model

Images