Method and device for generating human face cartoon based on deep learning

Abstract

The invention discloses a method and a device for generating a human face cartoon based on deep learning, wherein the method comprises the following steps: acquiring a personal image, and calibrating77 facial feature points; performing re-estimation and rotation operations to obtain an image of the human face with a positive standard posture; constructing a rectangular frame of the human face according to the image; zooming for three times in an equal ratio to perform cropping, and constructing a human head image block; at the same time, cropping an image corresponding to the rectangular frame of the human face, performing naturalization processing, sequentially feeding to a pre-trained human face and facial feature multi-label classification deep convolutional neural prediction network,and obtaining the gender and facial feature attribute labels of the human face; according to the images corresponding to the gender and facial feature attribute labels of the human face, the human head image block and the rectangular frame of the human face, selecting corresponding materials from a cartoon material library, performing corresponding deformation processing, and obtaining cartoon materials of the facial features, the hair and the human skin; then, performing splicing processing; finally, obtaining a cartoon image. In the embodiment of the present invention, intelligent drawing of human face cartoon can be realized, and any manual assistance is not required.

Description

Technical field [0001] The invention relates to the technical field of computer non-realistic rendering, in particular to a method and device for generating facial cartoons based on deep learning. Background technique [0002] In recent years, with the development of computer image processing technology, more and more facial comic generation applications such as Face Meng, Moman Camera, and FaceU have entered our lives. They all use computer image processing technology to perform real portraits. Rendering to generate vivid comic images of characters. Generally speaking, traditional comic generation systems that need to generate facial comic images can be divided into two categories. The first type is comic generation systems that rely on a combination of materials to create comics. A typical application example is the mobile phone application "face cute". Another type of comic generation device is based on computer image processing technology. It directly modifies real-life port...

AI Technical Summary

Problems solved by technology

For example, the technical shortcomings of "Face Moe": (1) A major factor in the generation of face caricatures is likeness, that is, there should be a certain degree of similarity between the generated caricatures and the target real face

[0004] (3) In Face Moe, the materials available for selection are limited, and the limited materials cannot meet the needs of everyone to create their own unique comic image

[0005] Another example is Moman Camera, its disadvantages: (1) The entire human face is regarded as an image, and the method of processing with computer image processing technology is used to generate a caricature face. The generated caricature face looks more like a gray person Face photo, relatively blunt, does not have the artistic effect of comics

[0006] (2) There is a mismatch between the real caricature face and other optional materials created by the artist, which makes the generated caricature look unnatural

[0007] (3) In order to match the comic face with other materials, the comic generation device needs to manually select the comic materials used to avoid a sense of disobedience, which leads to the extremely limited comic styles supported by Moman Camera

Images