US 11,842,468 B2
Synthesizing digital images utilizing image-guided model inversion of an image classifier
Pei Wang, Irvine, CA (US); Yijun Li, Seattle, WA (US); Jingwan Lu, Santa Clara, CA (US); and Krishna Kumar Singh, San Jose, CA (US)
Assigned to Adobe Inc., San Jose, CA (US)
Filed by Adobe Inc., San Jose, CA (US)
Filed on Feb. 18, 2021, as Appl. No. 17/178,681.
Prior Publication US 2022/0261972 A1, Aug. 18, 2022
Int. Cl. G06K 9/00 (2022.01); G06T 5/50 (2006.01); G06N 3/04 (2023.01); G06V 10/75 (2022.01); G06F 18/22 (2023.01); G06F 18/24 (2023.01)
CPC G06T 5/50 (2013.01) [G06F 18/22 (2023.01); G06F 18/24 (2023.01); G06N 3/04 (2013.01); G06V 10/751 (2022.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/20221 (2013.01); G06V 10/759 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A non-transitory computer readable storage medium comprising instructions that, when executed by at least one processor, cause a computing device to:
synthesize a digital image based on a target image providing an image guide by:
generating a first feature set comprising feature vectors of an initial image via a plurality of separate neural network layers of an image classifier at a plurality of resolutions;
generating a second feature set comprising feature vectors of a target image via the plurality of separate neural network layers of the image classifier at the plurality of resolutions;
constraining, at the plurality of separate neural network layers of the image classifier at the plurality of resolutions, the feature vectors of the first feature set relative to corresponding feature vectors of the second feature set; and
constraining, utilizing a patch-based discriminator, first image patches of the initial image relative to second image patches of the target image.