US 12,462,545 B2
Controllable neural networks or other controllable machine learning models
Tien Cheng Bau, Irvine, CA (US)
Assigned to Samsung Electronics Co., Ltd., Suwon-si (KR)
Filed by Samsung Electronics Co., Ltd., Suwon-si (KR)
Filed on Jul. 21, 2022, as Appl. No. 17/870,011.
Claims priority of provisional application 63/230,435, filed on Aug. 6, 2021.
Prior Publication US 2023/0040176 A1, Feb. 9, 2023
Int. Cl. G06V 10/82 (2022.01); G06N 3/045 (2023.01); G06T 3/4053 (2024.01); G06V 10/26 (2022.01)
CPC G06V 10/82 (2022.01) [G06N 3/045 (2023.01); G06T 3/4053 (2013.01); G06V 10/26 (2022.01)] 29 Claims
OG exemplary drawing
 
1. A method comprising:
obtaining, using at least one processor of an electronic device, a machine learning model trained to perform at least one image processing operation on image input data and to generate image output data over at least one range of values associated with one or more control variables, the machine learning model trained using a weighted combination of a plurality of loss functions including at least a first loss function corresponding to content loss, a second loss function corresponding to perceptual loss, and a third loss function corresponding to adversarial loss, wherein the one or more control variables relate to weighting of each of the plurality of loss functions within the weighted combination;
providing, using the at least one processor, specified image input data to the machine learning model;
providing, using the at least one processor, one or more specified values of the one or more control variables to the machine learning model, the one or more specified values of the one or more control variables within the at least one range of values; and
performing inferencing using the machine learning model to process the specified image input data using the at least one image processing operation and generate specified image output data, wherein the inferencing is controlled based on the one or more specified values of the one or more control variables.