US 12,423,959 B2
Method of data augmentation and non-transitory computer readable media
Chao-Hsun Yang, Hsinchu (TW); Chun-Chang Wu, Hsinchu (TW); and Shih-Tse Chen, Hsinchu (TW)
Assigned to Realtek Semiconductor Corporation, Hsinchu (TW)
Filed by Realtek Semiconductor Corporation, Hsinchu (TW)
Filed on May 5, 2022, as Appl. No. 17/662,077.
Claims priority of application No. 110128791 (TW), filed on Aug. 4, 2021.
Prior Publication US 2023/0041693 A1, Feb. 9, 2023
Int. Cl. G06V 10/774 (2022.01); G06V 10/25 (2022.01)
CPC G06V 10/7747 (2022.01) [G06V 10/25 (2022.01)] 18 Claims
OG exemplary drawing
 
1. A method of data augmentation, comprising:
selecting an original image from an original dataset comprising label data configured to indicate a labeled area of the original image, wherein the labeled area is a partial area of the original image;
selecting a first part of the labeled area with a first truncated object as a first target image;
generating a first sample image according to the first target image, comprising:
setting a height of the first sample image to a height of the first target image adding a first value randomly determined; and
setting a width of the first sample image to a width of the first target image adding a second value randomly determined,
wherein the first sample image comprises the first target image and a first border pattern different from the first target image, and the first part of the labeled area is free from comprising at least part of the first border pattern; and
incorporating the first sample image into a sample dataset, wherein the sample dataset is configured to be inputted to a machine learning model.