US 12,423,822 B2
Method and apparatus with optimization and prediction for image segmentation
Pingjun Li, Beijing (CN); Yi Zhou, Beijing (CN); Byung In Yoo, Seoul (KR); Hana Lee, Suwon-si (KR); and Hui Zhang, Beijing (CN)
Assigned to Samsung Electronics Co., Ltd., Suwon-si (KR)
Filed by Samsung Electronics Co., Ltd., Suwon-si (KR)
Filed on Dec. 10, 2021, as Appl. No. 17/547,767.
Claims priority of application No. 202011457838.5 (CN), filed on Dec. 11, 2020; and application No. 10-2021-0135124 (KR), filed on Oct. 12, 2021.
Prior Publication US 2022/0189031 A1, Jun. 16, 2022
Int. Cl. G06T 7/11 (2017.01); G06N 3/045 (2023.01); G06T 7/187 (2017.01)
CPC G06T 7/11 (2017.01) [G06T 7/187 (2017.01); G06N 3/045 (2023.01)] 14 Claims
OG exemplary drawing
 
1. A processor-implemented method, comprising:
in response to a probability that a pixel of an input image belongs to a plurality of category areas being greater than a threshold value, determining a category of the pixel to be a category corresponding to the plurality of category areas using an optimized second pixel classifier, wherein the second pixel classifier is optimized using a new single category image determined by removing a first category of the category areas from the single category image; and
in response to the probability being less than or equal to the threshold value,
determining a probability that the pixel belongs to a preset category, for each of a plurality of preset categories; and
determining the category of the pixel to be a category determined based on the probabilities determined for each of the preset categories.