US 11,854,244 B2
Labeling techniques for a modified panoptic labeling neural network
Sohrab Amirghodsi, Seattle, WA (US); Zhe Lin, Fremont, CA (US); Yilin Wang, Sunnyvale, CA (US); Tianshu Yu, Tempe, AZ (US); Connelly Barnes, Seattle, WA (US); and Elya Shechtman, Seattle, WA (US)
Assigned to ADOBE INC., San Jose, CA (US)
Filed by Adobe Inc., San Jose, CA (US)
Filed on Oct. 20, 2022, as Appl. No. 18/048,311.
Application 18/048,311 is a continuation of application No. 15/930,539, filed on May 13, 2020, granted, now 11,507,777.
Prior Publication US 2023/0079886 A1, Mar. 16, 2023
This patent is subject to a terminal disclaimer.
Int. Cl. G06V 10/75 (2022.01); G06F 17/18 (2006.01); G06N 3/08 (2023.01); G06N 20/00 (2019.01); G06V 10/82 (2022.01); G06F 18/214 (2023.01); G06F 18/22 (2023.01); G06F 18/211 (2023.01); G06F 18/213 (2023.01); G06V 10/74 (2022.01); G06V 10/771 (2022.01); G06V 10/774 (2022.01); G06V 20/70 (2022.01)
CPC G06V 10/757 (2022.01) [G06F 17/18 (2013.01); G06F 18/211 (2023.01); G06F 18/213 (2023.01); G06F 18/214 (2023.01); G06F 18/22 (2023.01); G06N 3/08 (2013.01); G06N 20/00 (2019.01); G06V 10/761 (2022.01); G06V 10/771 (2022.01); G06V 10/774 (2022.01); G06V 10/82 (2022.01); G06V 20/70 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A method of generating panoptic labeling for graphical digital images, the method comprising:
generating a modified input image based on an input image and a mask instance;
determining a subset of a set of labels determined from a group of images, each label in the subset representing a respective category of a respective object depicted in the input image;
determining, by a modified panoptic labeling neural network and based on a distance between a mask pixel from the mask instance and an object pixel from the modified input image, a probability of the mask pixel having a label from the subset, the object pixel having the label from the subset;
generating, by the modified panoptic labeling neural network, a mask label for the mask pixel based on the probability; and
providing the mask label to a digital graphics editing system.