EA received 1 granted patent in Graphics.
The patent describes a system that replaces traditional physics-based cloth simulation with a neural network approach, where the network predicts clothing mesh deformations and normal maps directly from character joint rotations to reduce computational costs during real-time rendering.
EA's single Graphics patent replaces the computationally expensive loop of physics simulation with a neural network that predicts cloth deformation in one pass. The system takes a character's skeletal pose as input and outputs compressed representations of the clothing mesh and normal map, encoded as PCA coefficients, without needing any information from previous frames. Training data comes from physics simulations run offline, where high-resolution cloth interactions are baked down into low-resolution meshes and normal maps that the network learns to reproduce. This stateless design avoids the frame-by-frame dynamics calculation that makes traditional cloth simulation costly in real-time rendering.
All data sourced from USPTO patent filings. Google Patents may take several weeks to index recent publications. If a link is unavailable, search for the patent number at USPTO Patent Public Search.