US 12,169,880 B2
Hierarchical model-based generation of images
Ondrej Texler, San Jose, CA (US); Dimitar Petkov Dinev, Sunnyvale, CA (US); Ankur Gupta, San Jose, CA (US); Hyun Jae Kang, Mountain View, CA (US); Anthony Sylvain Jean-Yves Liot, San Jose, CA (US); Siddarth Ravichandran, Santa Clara, CA (US); and Sajid Sadi, San Jose, CA (US)
Assigned to SAMSUNG ELECTRONICS CO., LTD., Gyeonggi-Do (KR)
Filed by SAMSUNG ELECTRONICS CO., LTD., Gyeonggi-Do (KR)
Filed on Oct. 17, 2022, as Appl. No. 17/967,868.
Claims priority of provisional application 63/349,289, filed on Jun. 6, 2022.
Prior Publication US 2023/0394715 A1, Dec. 7, 2023
Int. Cl. G06T 11/00 (2006.01); G06T 7/70 (2017.01); G06T 13/00 (2011.01)
CPC G06T 11/00 (2013.01) [G06T 7/70 (2017.01); G06T 13/00 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/30201 (2013.01)] 16 Claims
OG exemplary drawing
 
1. A computer-implemented method, comprising:
generating, by computer hardware using a first neural network model, a first region of an image based on a first set of one or more data modalities provided to the first neural network model as input;
providing the first region of the image to a second neural network model as input; and
generating, by the computer hardware using a second neural network model, a second region of the image based on a second set of one or more data modalities provided to the second neural network model as additional input, wherein the first set of one or more data modalities differs from the second set of one or more data modalities, and wherein the second region of the image shares a boundary with at least a portion of the first region of the image;
wherein the second neural network model includes a skip connection that directly links non-sequential layers of the second neural network model, and wherein the skip connection configures the second neural network model to output the first region unmodified and positioned within the second region.