US 11,941,869 B2
Method, device, and computer program product for data augmentation
Zijia Wang, WeiFang (CN); Wenbin Yang, Shanghai (CN); and Zhen Jia, Shanghai (CN)
Assigned to Dell Products L.P., Round Rock, TX (US)
Filed by Dell Products L.P., Round Rock, TX (US)
Filed on Feb. 21, 2022, as Appl. No. 17/676,426.
Claims priority of application No. 202210072009.8 (CN), filed on Jan. 21, 2022.
Prior Publication US 2023/0237780 A1, Jul. 27, 2023
Int. Cl. G06N 3/126 (2023.01); G06V 10/74 (2022.01); G06V 10/774 (2022.01)
CPC G06V 10/7747 (2022.01) [G06N 3/126 (2013.01); G06V 10/761 (2022.01)] 18 Claims
OG exemplary drawing
 
1. A method for data augmentation, comprising:
generating a group of candidate images based on a target image by using a first instance of a genetic algorithm, the first instance of the genetic algorithm being based on a thermodynamic genetic algorithm (TDGA) model, the TDGA model being configured to apply one or more operations of a set of predetermined image processing operations during each evolution process;
determining multiple augmented images from the group of candidate images based on free energy of the group of candidate images, the multiple augmented images being determined as belonging to the same classification with the target image;
utilizing a contrastive learning model to generate at least first and second feature representations of respective first and second different types based on the multiple augmented images;
utilizing a second instance of a genetic algorithm to determine respective weightings for the first and second feature representations of respective first and second different types; and
generating at least one saliency map based on the first and second feature representations and their respective weightings.