Real-Time Clothing Synthesis via Pose-Shape Factorization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current virtual clothes-fitting technologies face challenges in generating real-time cloth animations for various poses and shapes, with physics-based methods being time-consuming and data-driven methods compromising animation quality by assuming linearity or locality, while existing learning-based methods require substantial computational resources for database construction.
Innovation Solution
An example-based real-time clothing synthesis method that efficiently generates clothing animations by factorizing clothing deformation into pose and shape components, using SMPL models, sensitivity-optimized rigging, and Linear Blend Skinning (LBS) to predict deformations for different poses and shapes, and reduces computational resources through pose clustering and region-based skinning weights.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If physics-based simulation methods are used to model nonlinear behavior of clothing, then highly realistic cloth simulations are achieved, but the simulation becomes time-consuming and difficult to generate real-time cloth animation
Solution Approach 1:
The patent pre-computes clothing deformation samples for various body poses and shapes offline before runtime. These pre-computed examples are stored in a database and retrieved during real-time virtual try-on, eliminating the need for time-consuming physics simulations during runtime while maintaining realistic deformation patterns.
Solution Approach 2:
The patent creates simplified copies of complex physics-based clothing deformations by capturing deformation patterns from physics simulations and storing them as example data. These copies can be quickly retrieved and applied without re-running the full physics simulation, achieving real-time performance while preserving realistic deformation characteristics.
2Productivity
If data-driven cloth animation methods assume linearity or locality to achieve real-time performance, then real-time animation is generated, but the animation quality is compromised
Solution Approach 1:
The patent retrieves and blends multiple clothing deformation examples that are locally similar to the target pose, rather than relying on a single linear interpolation. This partial use of multiple examples provides better accuracy while maintaining real-time performance through efficient database querying and blending operations.
3Adaptability or versatility
If existing learning-based methods train models to predict deformation under various poses and shapes, then the try-on effect of same clothing under different postures and body types is predicted, but exhaustive database construction requires substantial computational resources
Solution Approach 1:
The patent performs exhaustive database construction offline before deployment, pre-computing clothing deformations for a comprehensive set of poses and shapes. This shifts the computational burden from runtime to offline preparation, enabling fast real-time predictions with minimal computational resources during actual virtual try-on operations.
Data Source
AI summary
The invention presents an example-based real-time clothing synthesis method, which includes the following steps: creating the 3D human model according to the input body pose parameters and the body shape parameters; generating 3D clothing deformation on the standard shape and the input pose through a clothing pose model; searching nearby clothing shape models according to the body pose parameters; for each nearby clothing shape model, generating corresponding 3D clothing mesh based on the body shape parameters, and then deforming the clothing mesh toward the input pose using Linear Blend Skinning (LBS) method; synthesizing the clothing deformation for the input body pose and shape using the clothing data obtained in above steps; resolving the interpenetrations between the clothing data and the human model. The invention can generate real-time clothing animation results with detailed wrinkles for various body poses and shapes with different clothing types on a commodity CPU, without resizing the cloth model. It has potential use in assessment of fitness in virtual try-on, as well as in the field of computer animation and video games.