Digital cloth real-time resolving method based on deep learning

Abstract

The invention provides a digital human cloth real-time calculation method based on deep learning. The method comprises the following steps: S1, training a neural network, and training an automatic encoder and a full-connection network to map skeleton information to codes of cloth grid vertexes; and S2, after the two neural networks are trained, giving a group of input skeleton information, firstlyobtaining codes of cloth grid vertexes through a full connection network, and then obtaining output cloth grid vertexes through a decoding link in an automatic encoder, wherein the output is a calculation result of the cloth. The grid vertex information of the cloth is calculated in real time, the problem of interpenetration of a figure model and a cloth model is effectively avoided, and the cloth manufacturing efficiency is further improved.

Description

technical field [0001] The invention belongs to the technical field of animation production, and in particular relates to a real-time calculation method for digital human cloth based on deep learning. Background technique [0002] In the current film and television animation production process, the animation data of clothing cloth is calculated by establishing a mathematical physical model to simulate the movement of the cloth in a real scene to calculate the final effect, that is, the three-dimensional coordinate value of the vertices of the cloth mesh in each frame. Modeling and simulation means that calculating the three-dimensional coordinates of the vertices of each frame consumes a lot of resources. Therefore, the current digital human cloth generally adopts the offline calculation method, that is, the cloth of each frame is calculated in advance according to the skeleton animation of the character. Grid vertex information, written into the cache, and called out when n...

AI Technical Summary

Problems solved by technology

[0003] Due to the nature of the calculation process of modeling simulation, the specific size and shape of the cloth must be changed according to the change of the character model, otherwise it will lead to the phenomenon that the cloth and the character model are interspersed, and manual refinement is required

[0004] After the cloth animation is solved, if the character skeleton animation is modified in the later stage, the corresponding cloth animation needs to be recalculated to generate a new cache, and the cloth production efficiency is low.

Images