MULTI-HAIRSTYLE REAL-TIME ANIMATION METHOD BASED ON NEURAL NETWORK INTERPOLATOR

WIPO Patent Application WO/2022/222058

A1

Disclosed in the present invention is a multi-hairstyle real-time animation method based on a neural network interpolator. By means of the method, parameterized reconstruction is performed on an original hairstyle, such that the original hairstyle is applicable to neural network learning; a neural network interpolator is pre-trained using full-simulation hair motion data; and in a real-time application stage, an animation of all hair is recovered by means of a small amount of hair simulated in real time and the neural network interpolator. By means of the present invention, a multi-hairstyle real-time animation can be realized; and the aim of a real-time animation of all hair is achieved by means of simulation of a small amount of guide hair in a low computing resource consumption and by means of a pre-trained neural network interpolator. By means of the present invention, with respect to the simulation of all hair, the computing resource consumption is reduced, and a multi-hairstyle situation can also be handled, thereby expanding the application range of the method.

ZHOU KUN (CN)
LYU QING (CN)
CHEN XIANG (CN)

PCT/CN2021/088665

October 27, 2022

April 21, 2021

Click for automatic bibliography generation  

UNIV ZHEJIANG (CN)
FACEUNITY TECH CO LTD (CN)

G06T17/00; G06T7/73

CN110930500A	2020-03-27
CN111583384A	2020-08-25

LYU QING, CHAI MENGLEI, CHEN XIANG, ZHOU KUN: "Real-Time Hair Simulation With Neural Interpolation", IEEE TRANSACTIONS ON VISUALIZATION AND COMPUTER GRAPHICS, IEEE, USA, vol. 28, no. 4, 1 April 2022 (2022-04-01), USA, pages 1894 - 1905, XP055979087, ISSN: 1077-2626, DOI: 10.1109/TVCG.2020.3029823

HANGZHOU QIUSHI PATENT OFFICE CO., LTD. (CN)

View/Download PDF      PDF Help

Previous Patent: OUT OF DISTRIBUTION SAMPLES REPORTING FOR NEURAL NETWORK OPTIMIZATION

Next Patent: TREATMENT APPARATUS