US 12,136,250 B2
Extracting attributes from arbitrary digital images utilizing a multi-attribute contrastive classification neural network
Khoi Pham, Fairfax, VA (US); Kushal Kafle, Boston, MA (US); Zhe Lin, Fremont, CA (US); Zhihong Ding, Fremont, CA (US); Scott Cohen, Sunnyvale, CA (US); and Quan Tran, San Jose, CA (US)
Assigned to Adobe Inc., San Jose, CA (US)
Filed by Adobe Inc., San Jose, CA (US)
Filed on May 27, 2021, as Appl. No. 17/332,734.
Prior Publication US 2022/0383037 A1, Dec. 1, 2022
Int. Cl. G06V 10/75 (2022.01); G06F 18/214 (2023.01); G06F 18/25 (2023.01); G06N 3/08 (2023.01)
CPC G06V 10/751 (2022.01) [G06F 18/214 (2023.01); G06F 18/25 (2023.01); G06N 3/08 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A non-transitory computer-readable medium storing instructions that, when executed by at least one processor, cause a computing device to:
generate an image-object feature map reflecting attributes from a digital image portraying an object utilizing an embedding neural network;
generate a localized object attention feature vector reflecting a segmentation prediction of the object portrayed in the digital image from the image-object feature map utilizing a localizer neural network; and
determine a plurality of attributes for the object portrayed within the digital image from a combination of the localized object attention feature vector and the image-object feature map utilizing a classifier neural network.