Motion Completion Model Training for Smooth Digital Human Animation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing motion completion methods using linear interpolation result in imprecise and inefficient motion frame insertion, leading to discontinuous and unsmooth motion sequences in digital or virtual humans.
Innovation Solution
A method for training a motion completion model that involves determining motion sub-sequences, performing masking processing, and updating model parameters based on differences between predicted and original motion frames, to improve precision and efficiency of motion completion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If linear interpolation is used for motion completion, then the process is simple and fast, but the precision of motion frames is low
Solution Approach 1:
The patent replaces the mechanical linear interpolation calculation method with a deep learning-based motion completion model. The model uses neural networks to predict missing motion frames, substituting the simple mathematical interpolation with an intelligent system that learns motion patterns from data, thereby achieving higher precision while managing complexity through automated learning.
Solution Approach 2:
The patent changes the fundamental parameter of motion completion from linear mathematical interpolation to non-linear neural network prediction. By transforming the completion method from a deterministic mathematical approach to a probabilistic learning-based approach, the system achieves superior motion frame precision while the model parameters are optimized through training data.
2Productivity
If linear interpolation is used for motion completion, then the computational process is fast, but the efficiency of motion completion is not high
Solution Approach 1:
The patent applies preliminary action by pre-training the motion completion model on extensive motion data before actual use. The model learns motion patterns, transitions, and temporal relationships in advance through training phases, so that during actual motion completion tasks, it can rapidly generate high-precision results without performing complex real-time calculations, thus improving both efficiency and precision.
Solution Approach 2:
The patent uses copying by training the model on numerous examples of motion sequences and their completions. The neural network learns to copy and generalize motion patterns from training data, enabling it to efficiently generate accurate completion frames by replicating learned motion dynamics rather than calculating each frame from scratch.
3Ease of manufacture
If linear interpolation is used, then the implementation is straightforward, but the motion frames obtained are not precise and efficiency is not high
Solution Approach 1:
The patent implements self-service by enabling the motion completion model to automatically learn and improve through training without requiring manual intervention for each completion task. The system serves itself by using training data to automatically adjust its parameters and optimize its performance, making the implementation process more efficient while maintaining ease of use through automated learning and adaptation.
Data Source
AI summary
This application provides a method for training a motion completion model performed by an electronic device. The method includes: obtaining a motion sequence sample, the motion sequence sample including at least three consecutive motion frames; determining at least one first motion sub-sequence sample from the motion sequence sample, the first motion sub-sequence sample having two second motion sub-sequence samples adjacent thereto; performing masking processing on the at least one first motion sub-sequence sample, to obtain a target motion sequence sample; performing motion completion processing on the target motion sequence sample through a motion completion model, to obtain a completing motion sequence; and updating a model parameter of the motion completion model based on a difference between the completing motion sequence and the at least one first motion sub-sequence sample, to obtain a trained motion completion model.