Pose Prediction Model for Runtime Game Animation Generation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing video game technologies face challenges in generating realistic animations for characters in various environments, particularly those that are difficult or unsafe to capture with motion capture data, such as underwater or alien environments.
Innovation Solution
A computer-implemented method and system that uses a pose prediction model to generate realistic character animations by applying character poses and virtual environment labels, allowing for the creation of animations in new environments without extensive motion capture data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If motion capture data is used to generate animations, then animation quality is improved, but storage requirements and data collection complexity increase significantly
Solution Approach 1:
The patent creates a virtual copy of the motion capture process by training a neural network model on motion capture data. Once trained, the model generates animation frames without requiring actual motion capture data, effectively copying the motion patterns learned during training and applying them to new scenarios, thereby reducing storage needs while maintaining animation quality
Solution Approach 2:
The patent transforms the animation generation process from storing complete motion capture datasets to storing a trained neural network model with learned parameters. This parameter transformation allows the system to generate diverse animations from a compact model representation, significantly reducing storage requirements while preserving animation quality
2Adaptability or versatility
If motion capture data is collected for all possible interactions, then animation coverage is improved, but data collection difficulty and time increase
Solution Approach 1:
The patent performs preliminary action by collecting motion capture data for a limited set of representative interactions during the training phase. The neural network model learns general motion patterns from this preliminary data, enabling it to generate animations for unlimited interactions without requiring additional data collection, thus saving time while maintaining comprehensive animation coverage
Solution Approach 2:
The trained neural network model serves as a universal animation generator that can handle multiple different interactions and environments. Instead of collecting separate motion capture data for each interaction type, the single trained model generalizes across diverse scenarios, achieving broad animation coverage without proportional increases in data collection time
3Adaptability or versatility
If more animation files are stored to cover various movements, then character movement variety is improved, but storage space requirements increase
Solution Approach 1:
The patent replaces the need to store multiple animation files with a single trained neural network model that can generate diverse animations on-demand. The model copies learned motion patterns from training data and generates new animation sequences dynamically, achieving high character movement variety without proportionally increasing storage space
Solution Approach 2:
The patent transitions from static pre-recorded animation files to dynamic on-the-fly animation generation. The neural network model dynamically generates appropriate animations based on real-time game state and character interactions, providing unlimited movement variety while using minimal storage space compared to storing all possible animation variants
Data Source
AI summary
Use of pose prediction models enables runtime animation to be generated for an electronic game. The pose prediction model can predict a character pose of a character based on joint data for a pose of the character in a previous frame. Further, by using environment data, it is possible to modify the prediction of the character pose based on a particular environment in which the character is location. Advantageously, the use of machine learning enables prediction of character movement in environment in which it is difficult or impossible to obtain motion capture data.