3D Character Generation via Mesh Segmentation and Assembly
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for generating three-dimensional (3D) characters are labor-intensive and resource-intensive, requiring significant human involvement and technical skill, making it challenging to produce a large number of unique and high-quality 3D characters efficiently, especially when thousands are needed for applications like gaming and film.
Innovation Solution
A computing device standardizes 3D models through scaling and reorientation, applies a base mesh with material IDs, extracts and assembles polygonal mesh pieces and texture map features to generate composite meshes and texture maps, and uses machine learning to ensure uniqueness, allowing for rapid creation of diverse 3D characters from a limited set of source models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If sculpting/modeling and blend-shapes/morphing methods are used to create unique 3D characters, then character uniqueness and detail quality are improved, but labor intensity and resource consumption increase significantly
Solution Approach 1:
The patent segments a 3D character model into multiple distinct components or regions (e.g., head, torso, limbs, facial features). Each segment can be independently selected, modified, or generated from a library of pre-defined options, enabling efficient assembly of unique characters without manual sculpting of entire models.
Solution Approach 2:
The patent utilizes parameter-based control to generate character variations. By adjusting parameters such as scale, orientation, material properties, and geometric transformations, the system can rapidly produce diverse character designs from base templates, maintaining high detail quality while automating the generation process.
2Manufacturing precision
If scans of human models are taken through LIDAR, photogrammetry, and hand scanning to create 3D characters, then character realism and detail are improved, but the number of required scans and human involvement increase considerably
Solution Approach 1:
The patent creates digital copies or representations of character components from a limited set of source scans. These copied elements are stored in libraries and can be reused, combined, and transformed to generate numerous unique characters without requiring new scans for each character, dramatically reducing data acquisition time.
Solution Approach 2:
The patent develops universal base models and component libraries that can serve multiple character generation purposes. A single scanned model can be decomposed into reusable components that function across different character designs, allowing the same source data to generate diverse character sets for various applications.
3Productivity
If a limited set of source models is used to generate thousands of unique 3D characters, then resource efficiency is improved, but character uniqueness and diversity become more challenging to achieve
Solution Approach 1:
The patent implements dynamic selection and combination mechanisms that adaptively assemble character components. Through algorithms that randomly or strategically combine different segments, materials, and transformations from the source library, the system generates diverse character variations while maintaining production efficiency.
Solution Approach 2:
The patent introduces additional dimensions of variation beyond the original source models by applying transformations in multiple parameter spaces (scaling, rotation, material assignment, geometric deformation). This allows the system to generate character diversity by exploring parameter combinations rather than requiring proportionally more source models.
Data Source
AI summary
Aspects described herein relate to three-dimensional (3D) characters and rapidly generating 3D characters from a plurality of 3D source models. In generating the 3D characters, one or more 3D source models may be standardized, applied to a base mesh with material ID assignments, and decomposed to isolate particular polygonal mesh pieces and/or texture map feature selections denoted by the material IDs of the base mesh. The disparate isolated polygonal mesh pieces and/or texture map feature selections may be assembled in a modular fashion to compose unique 3D characters unlike any of those of the one or more 3D models. The 3D characters may be further refined through the addition of features and/or accessories, and may also be processed through machine learning algorithms to further ascertain uniqueness.