Three-dimensional face model training method and device, electronic equipment and readable storage medium

Abstract

The invention provides a three-dimensional face model training method and device, electronic equipment and a readable storage medium, after a face image is obtained, a face region image in the face image is cut, and the face region image comprises a plurality of face feature points. Constructing a virtual perspective camera, obtaining perspective 3D position information and 2D position information of each face feature point in the face region image in the space of the virtual perspective camera, and training the constructed neural network model by using the face region image carrying the perspective 3D position information and the 2D position information; and a three-dimensional face model which can be used for detecting the three-dimensional information of the face feature points is obtained. According to the scheme, the three-dimensional face model capable of detecting the three-dimensional information of the face feature points is trained, accurate detection of the 3D key points is realized, and model training is performed in combination with perspective 3D position information and 2D position information, so that the trained model has a relatively good effect on face feature point alignment.

Description

technical field [0001] The present application relates to the technical field of image processing, in particular, to a three-dimensional face model training method, device, electronic equipment and readable storage medium. Background technique [0002] With the rise of live broadcast entertainment and short videos, users’ demand for facial special effects such as face beautification and automatic makeup is also increasing. The 3D face reconstruction technology based on deep learning has also made great breakthroughs as the basis for realizing these functions. Most of the face 3D reconstruction methods used in the prior art are based on the 3D deformable Basel face model, that is, the 3D key point information of the face is reconstructed by predicting the expression base vector and shape base vector of the face . Although the spatial constraints of this method are relatively strong, it is very difficult to obtain the expression base vector and shape base vector. At present, ...

AI Technical Summary

Problems solved by technology

Although the spatial constraints of this method are relatively strong, it is very difficult to obtain the expression basis vector and shape basis vector. At present, it is generally calculated by least squares fitting and calculation of the 3D key points and 2D key points of the face. There are certain problems in the method itself, and it is difficult to obtain an accurate solution

Images