20240037827. RESOLVING GARMENT COLLISIONS USING NEURAL NETWORKS simplified abstract (Adobe Inc.)
Contents
RESOLVING GARMENT COLLISIONS USING NEURAL NETWORKS
Organization Name
Inventor(s)
Yangtuanfeng Wang of London (GB)
Qingyang Tan of Greenbelt MD (US)
Duygu Ceylan Aksit of London (GB)
RESOLVING GARMENT COLLISIONS USING NEURAL NETWORKS - A simplified explanation of the abstract
This abstract first appeared for US patent application 20240037827 titled 'RESOLVING GARMENT COLLISIONS USING NEURAL NETWORKS
Simplified Explanation
The disclosed patent application describes a system and method for using machine learning models to deform three-dimensional garments based on the motion of a character's body, while also handling collisions.
- The system receives an input that includes parameters defining the shape and pose of a character's body, as well as garment parameters.
- A first neural network generates a set of garment vertices that define the deformations of the garment based on the input.
- A second neural network determines if any of the garment vertices penetrate the character's body.
- A third neural network modifies the positions of the penetrating garment vertices to move them outside of the character's body.
Potential applications of this technology:
- Virtual try-on: This technology can be used in virtual shopping experiences to allow customers to see how garments deform and fit on their virtual avatars.
- Animation and gaming: It can be used to create realistic clothing deformations and movements in animated characters or video game avatars.
- Virtual reality: This technology can enhance the immersion in virtual reality experiences by providing realistic garment deformations based on the user's body motion.
Problems solved by this technology:
- Accurate garment deformations: Traditional methods for simulating garment deformations may not accurately capture the complex interactions between the character's body and the garment. This technology uses machine learning models to improve the accuracy of the deformations.
- Collision handling: Handling collisions between the garment and the character's body is a challenging problem. This technology addresses this issue by modifying the garment vertices to avoid penetration.
Benefits of this technology:
- Realism: By accurately simulating garment deformations based on body motion, this technology enhances the realism of virtual experiences.
- Efficiency: The use of machine learning models allows for faster and more efficient computation of garment deformations compared to traditional simulation methods.
- Customization: This technology can be used to create personalized virtual try-on experiences, where garments are deformed based on the user's specific body shape and pose.
Original Abstract Submitted
embodiments are disclosed for using machine learning models to perform three-dimensional garment deformation due to character body motion with collision handling. in particular, in one or more embodiments, the disclosed systems and methods comprise receiving an input, the input including character body shape parameters and character body pose parameters defining a character body, and garment parameters. the disclosed systems and methods further comprise generating, by a first neural network, a first set of garment vertices defining deformations of a garment with the character body based on the input. the disclosed systems and methods further comprise determining, by a second neural network, that the first set of garment vertices includes a second set of garment vertices penetrating the character body. the disclosed systems and methods further comprise modifying, by a third neural network, each garment vertex in the second set of garment vertices to positions outside the character body.
