	US 12,482,224 B2
	Electronic device and method with spatio-temporal self-similarity consideration
Minsu Cho, Pohang-si (KR); Heeseung Kwon, Pohang-si (KR); Manjin Kim, Pohang-si (KR); and Suha Kwak, Pohang-si (KR)
Assigned to Samsung Electronics Co., Ltd., Suwon-si (KR); and POSTECH Research and Business Development Foundation, Pohang-si (KR)
Filed by Samsung Electronics Co., Ltd., Suwon-si (KR); and POSTECH Research and Business Development Foundation, Pohang-si (KR)
Filed on Oct. 31, 2022, as Appl. No. 17/977,449.
Claims priority of application No. 10-2021-0164497 (KR), filed on Nov. 25, 2021.
Prior Publication US 2024/0153238 A1, May 9, 2024
Int. Cl. G06V 10/74 (2022.01); G06T 7/246 (2017.01); G06T 7/73 (2017.01); G06V 10/62 (2022.01); G06V 10/75 (2022.01); G06V 10/764 (2022.01); G06V 10/77 (2022.01); G06V 10/82 (2022.01); G06V 40/20 (2022.01)

CPC G06V 10/761 (2022.01) [G06T 7/248 (2017.01); G06T 7/74 (2017.01); G06V 10/62 (2022.01); G06V 10/751 (2022.01); G06V 10/759 (2022.01); G06V 10/764 (2022.01); G06V 10/7715 (2022.01); G06V 10/82 (2022.01); G06V 40/20 (2022.01); G06T 2207/10016 (2013.01); G06T 2207/20084 (2013.01)]

24 Claims

1. An electronic device comprising:

a processor; and

a memory comprising instructions that when executed by the processor configure the processor to:

generate, for respective positions of a video feature map corresponding to an input video, a spatio-temporal self-similarity (STSS) tensor comprising a spatial self-similarity map and spatial cross-similarity maps, wherein the spatial self-similarity map comprises similarities between a position in a frame of the input video and neighboring regions in the frame, and the spatial cross-similarity maps comprise similarities between the position in the frame and respectively corresponding positions in other frames of the input video;

generate STSS feature vectors from the STSS tensor;

generate an STSS feature map based on the STSS feature vectors; and

perform an inference on the input video based on the STSS feature map.