US 12,340,586 B2
Video semantic segmentation with bidirectional feature propagation and attention-based feature correction
Razieh Kaviani Baghbaderani, Rockville, MD (US); Yuanxin Li, San Diego, CA (US); and Shuangquan Wang, San Diego, CA (US)
Assigned to SAMSUNG ELECTRONICS CO., LTD., (KR)
Filed by Samsung Electronics Co., Ltd., Suwon-si (KR)
Filed on Jun. 17, 2022, as Appl. No. 17/843,949.
Claims priority of provisional application 63/254,051, filed on Oct. 8, 2021.
Prior Publication US 2023/0115081 A1, Apr. 13, 2023
Int. Cl. G06V 20/40 (2022.01)
CPC G06V 20/49 (2022.01) [G06V 20/46 (2022.01)] 13 Claims
OG exemplary drawing
 
1. A method to provide video semantic segmentation, the method comprising:
semantically segmenting, by a first image semantic-segmentation network, a first frame of a sequence of video frames to obtain at least one first semantic feature of the first frame;
semantically segmenting, by a second image semantic-segmentation network, a second frame of the sequence of video frames to obtain at least one second semantic feature of the second frame, the second frame being subsequent to the first frame;
semantically segmenting, by a third image semantic-segmentation network, a third frame of the sequence of video frames to obtain at least one third semantic feature of the third frame, the third frame being subsequent to both the first frame and the second frame according to a predetermined number of consecutive frames; and
generating, by a combining network, at least one fourth semantic feature of the second frame based on the at least one first semantic feature, the at least one second semantic feature and the at least one third semantic feature.