US 12,333,731 B2
Transformer for efficient image segmentation
Yilin Wang, Sunnyvale, CA (US); Chenglin Yang, Towson, MD (US); Jianming Zhang, Fremont, CA (US); He Zhang, Santa Clara, CA (US); Zijun Wei, San Jose, CA (US); and Zhe Lin, Clyde Hill, WA (US)
Assigned to ADOBE INC., San Jose, CA (US)
Filed by ADOBE INC., San Jose, CA (US)
Filed on Jun. 10, 2022, as Appl. No. 17/806,314.
Prior Publication US 2023/0401717 A1, Dec. 14, 2023
Int. Cl. G06K 9/00 (2022.01); G06T 7/11 (2017.01); G06V 10/778 (2022.01); G06V 10/82 (2022.01); G06V 20/70 (2022.01)
CPC G06T 7/11 (2017.01) [G06V 10/778 (2022.01); G06V 10/82 (2022.01); G06V 20/70 (2022.01); G06T 2207/20021 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01)] 20 Claims
OG exemplary drawing
 
1. A method comprising:
receiving an image depicting an object;
generating an atrous query matrix for the image by performing an atrous convolution operation based on a plurality of dilation rates;
generating image features for the image by performing an atrous self-attention operation based on the atrous query matrix; and
generating label data that identifies the object based on the image features.