A 3D face modeling method and printing device based on video stream and face multi-attribute matching

Abstract

The invention discloses a three-dimensional face modeling method and a printing device based on video stream and face multi-attribute matching, the method includes: establishing a general three-dimensional face model library; learning a deep neural network through pre-trained multi-task Perform face detection and face key point information extraction; use pre-trained multi-task learning deep neural network to analyze and predict face attributes, combine face key point data and face attribute information with the general 3D face model The library performs rough registration to obtain the most common face model that is closest to real-time acquisition. The printing device includes: an input unit, a feature point marking unit, a general three-dimensional face model library unit, a face multi-attribute classification unit, a face key point optimization unit, a texture refinement unit and a printout unit. The invention has the advantages of high precision, simple implementation, good user friendliness, high degree of automation and the like.

Description

technical field [0001] The invention relates to a three-dimensional modeling technology of a human face, in particular to a three-dimensional human face modeling method and a printing device based on video stream and multi-attribute matching of a human face. Background technique [0002] Many researchers make full use of the computer's excellent performance in image processing to simulate and demonstrate 3D face models and have made great achievements. Facial animation has evolved from traditional keyframe techniques to performance-driven techniques. From videophone to game entertainment, from multi-channel user interface to virtual reality, the technology of facial modeling and animation is everywhere. At the same time, people have increasingly strict requirements on the effect and quality of 3D face modeling using computers, and are no longer satisfied with 3D faces and their animations with stiff movements, dull expressions, and monotonous backgrounds. How to improve th...

AI Technical Summary

Problems solved by technology

[0005] The methods based on 3D scanners often have the problems of poor versatility and flexibility. In addition, the amount of data in the modeling process is large, the operation is relatively complicated, and the cost of the hardware equipment is expensive, and the calculation complexity is too high.

[0006] The modeling method based on a single face image is difficult to obtain a good modeling effect due to the complex calculation process, which leads to long calculation time and large deviation of calculation results.

[0007] Techniques based on fitting or reconstruction generally require user cooperation and are less user-friendly

For example, manual marking of key points, user registration before use, and the need to maintain no expression or fixed expression during modeling, etc., cannot accurately simulate real-time user expressions; at the same time, appearance and posture have a great impact on the accuracy of expression parameters. There is a certain bottleneck in mold accuracy

Images