US 12,430,726 B2
Semantic image extrapolation method and apparatus
Jaeseob Shin, Seoul (KR); Sungul Ryoo, Seoul (KR); Sehoon Son, Seoul (KR); Hyeongduck Kim, Gyeonggi-do (KR); and Hyosong Kim, Seoul (KR)
Assigned to PIXTREE CO., LTD., (KR)
Appl. No. 17/789,167
Filed by PIXTREE CO., LTD., Seoul (KR)
PCT Filed Dec. 18, 2020, PCT No. PCT/KR2020/018684
§ 371(c)(1), (2) Date Jun. 24, 2022,
PCT Pub. No. WO2021/133001, PCT Pub. Date Jul. 1, 2021.
Claims priority of application No. 10-2019-0174883 (KR), filed on Dec. 26, 2019.
Prior Publication US 2023/0051832 A1, Feb. 16, 2023
Int. Cl. G06T 5/77 (2024.01); G06T 5/50 (2006.01); G06T 7/11 (2017.01); G06T 7/13 (2017.01); G06V 10/764 (2022.01); G06V 20/70 (2022.01)
CPC G06T 5/77 (2024.01) [G06T 5/50 (2013.01); G06T 7/11 (2017.01); G06T 7/13 (2017.01); G06V 10/764 (2022.01); G06V 20/70 (2022.01); G06T 2207/20081 (2013.01)] 10 Claims
OG exemplary drawing
 
1. A semantic image extrapolation method comprising:
receiving an input image (I);
generating a segmentation map (S ) using a first artificial intelligence model learned in advance on the basis of deep learning from the input image (I);
generating an extrapolated segmentation map (S E) on the basis of the segmentation map (S ) using a second artificial intelligence model;
generating a padded image (IP) including a region for image-extension on the basis of the input image (I); and
generating an extrapolated image (I E) by combining the padded image (IP) and the extrapolated segmentation map (S E), using a third artificial intelligence model,
wherein in the generation of the extrapolated segmentation map (S E), instead of keeping the empty region in the form of a mask, a padded segmentation map (S P) in which an empty region is filled through interpolation is first generated, and the padded segmentation map (S P) is generated as the extrapolated segmentation map (S E) using the second artificial intelligence model.