US 12,450,791 B1
Method for generating deformed images for anomaly detection and computing device for performing the same
Jong Won Choi, Gyeongsangnam-do (KR); Min Gyu Lee, Seoul (KR); Do Hee Kim, Seoul (KR); and Yu Jeong Oh, Gyeonggi-do (KR)
Assigned to CHUNG ANG UNIVERSITY INDUSTRY ACADEMIC COOPERATION FOUNDATION, Seoul (KR)
Filed by CHUNG ANG UNIVERSITY INDUSTRY ACADEMIC COOPERATION FOUNDATION, Seoul (KR)
Filed on Feb. 20, 2025, as Appl. No. 19/058,168.
Claims priority of application No. 10-2024-0118846 (KR), filed on Sep. 2, 2024.
Int. Cl. G06T 11/00 (2006.01); G06F 17/10 (2006.01); G06F 40/40 (2020.01); G06N 3/08 (2023.01); G06T 9/00 (2006.01)
CPC G06T 11/001 (2013.01) [G06F 40/40 (2020.01); G06T 9/00 (2013.01)] 13 Claims
OG exemplary drawing
 
1. A method for generating deformed images performed on a computing device that includes one or more processors and a memory storing one or more programs executed by the one or more processors, the method comprising:
receiving an original image including a non-defective object and augmenting the original image into a plurality of images;
receiving the plurality of augmented original images and generating a deformed image from each of the augmented original images;
receiving a name of an object included in the original image as text and generating an optimal prompt based on the name of the object; and
determining whether to store the deformed image in a non-defective data pool based on the optimal prompt and the plurality of generated deformed images,
wherein the generating of the deformed image includes:
inputting the augmented original images into a first encoder to extract respective embedding vectors;
assigning index values to the embedding vectors and storing the embedding vectors in a code book;
calculating an average for each of the embedding vectors extracted from the first encoder;
calculating a variance of the embedding vectors in the code book;
calculating a quantized vector for each augmented original image by multiplying the calculated variance by noise and then adding the calculated average to the multiplication result; and
inputting the quantized vector for each augmented original image into a decoder to generate a deformed image for each augmented original image.