Action completion model training method, device, completion method, equipment and medium

Abstract

This application provides a training method, device, completion method, equipment, and medium for an action completion model. The training method of the model includes: collecting a large number of raw action numbers of human body movements, performing preprocessing to obtain a data set, and using the data set to analyze the action Completing the neural network model for training and optimization to obtain a trained motion-completing neural network model. Based on the well-trained motion completion neural network model, it can predict and calculate the data that automatically generates a large number of motion transition frames, which is suitable for scenes with long generation time, large number of transition frames, and complex motions, and the generated transition frames are realistic, reducing the original The labor cost required for the interpolation scheme improves the project quality and shortens the project cycle. The model training device, action completion method, computer equipment, and computer-readable storage medium provided by the present application all have the above-mentioned beneficial effects.

Description

Technical field [0001] The present application relates to the field of training the neural network model, particularly to a method of operation of a full complement of training model, means completion methods, apparatus and media. Background technique [0002] In the three-dimensional animation or games and other applications, the three-dimensional virtual image of the animation is driven by data, forming a walk, run, jump, dance and other movements. Animation data can be handmade by animators, or equipment obtained through motion capture. Motion capture animation data capture devices often can not be used directly, often not satisfied with the action in some segments, but also hand-animator changes. Therefore, in both cases, you need to manually edit the animator. [0003] When animators animation software, may edit or modify several key frames, and after the designated number of associated key frame animation software built by mathematical interpolation method, the purpose of g...

AI Technical Summary

Problems solved by technology

However, this method is only suitable for action transitions with short generation time, few transition frames, and simple actions

However, in the actual production process, an animator has to edit or modify a large number of key frames. If the number of key frames is too small, the time interval is too short, or the action is too complicated, the interpolation method will get an unreal and unreasonable transition sequence.

Using the existing solution will rely heavily on the skills and experience of the animators, thereby increasing the labor cost; due to the unreasonable transition sequences that are prone to occur, and the uneven technical level of the animators and other personnel, the quality of the project will also be affected ; At the same time, due to the need to debug and edit a large number of keyframes, it will greatly increase the project cycle

Images