US 11,682,213 B2
Method and device for training image analysis neural network model and intelligent image capturing apparatus employing the same
Hyun Jin Yoon, Daejeon (KR); and Mi Kyong Han, Daejeon (KR)
Assigned to ELECTRONICS AND TELECOMMUNICATIONS RESEARCH INSTITUTE, Daejeon (KR)
Filed by Electronics and Telecommunications Research Institute, Daejeon (KR)
Filed on Apr. 20, 2021, as Appl. No. 17/235,537.
Claims priority of application No. 10-2020-0047932 (KR), filed on Apr. 21, 2020.
Prior Publication US 2021/0326654 A1, Oct. 21, 2021
Int. Cl. G06K 9/00 (2022.01); G06V 20/52 (2022.01); G06N 3/08 (2023.01); G06F 18/214 (2023.01); G06F 18/2415 (2023.01); G06F 18/2431 (2023.01); G06V 10/764 (2022.01); G06V 10/82 (2022.01); G06V 20/70 (2022.01)
CPC G06V 20/52 (2022.01) [G06F 18/2155 (2023.01); G06F 18/2415 (2023.01); G06F 18/2431 (2023.01); G06N 3/08 (2013.01); G06V 10/764 (2022.01); G06V 10/82 (2022.01); G06V 20/70 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A method of training a neural network model for use in an analysis of a captured image, comprising:
providing the neural network model with an unlabeled first image and a labeled second image to obtain a first reconstructed image and a second reconstructed image specific to an image environment;
providing the neural network model with the unlabeled first image, the labeled second image, an image label of the labeled second image, and a domain label indicating to which category of a first category and a second category an image belongs to obtain a second prediction label for the labeled second image and a domain prediction label for the domain label;
calculating at least one loss function using the unlabeled first image, the labeled second image, the second prediction label for the labeled second image, and the domain prediction label for the domain label; and
updating parameters of the neural network model such that the at least one loss function is minimized.