US 12,406,422 B2
3D digital avatar generation from a single or few portrait images
Koki Nagano, Playa Vista, CA (US); and Jaewoo Seo, Kirkland, WA (US)
Assigned to NVIDIA Corporation, Santa Clara, CA (US)
Filed by NVIDIA Corporation, Santa Clara, CA (US)
Filed on Mar. 16, 2023, as Appl. No. 18/185,217.
Claims priority of provisional application 63/320,864, filed on Mar. 17, 2022.
Prior Publication US 2023/0298243 A1, Sep. 21, 2023
Int. Cl. G06T 13/40 (2011.01); G06T 13/80 (2011.01)
CPC G06T 13/40 (2013.01) [G06T 13/80 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method for generating a digital avatar from a two-dimensional (2D) input image in accordance with a first machine learning (ML) model and a second machine learning (ML) model, the method comprising:
determining a latent code for the second ML model, based at least on the 2D input image, the determining the latent code for the second ML model comprising:
executing a GAN inversion optimization algorithm for the first ML model to determine a latent code for the first ML model; and
executing a GAN inversion optimization algorithm for the second ML model to determine the latent code for the second ML model; and
generating, via the second ML model, three-dimensional (3D) data and color data representing the digital avatar in accordance with the latent code for the second ML model.