US 12,243,293 B2
System and method for generating training data sets for specimen defect detection
Anuj Doshi, Long Island City, NY (US); Jonathan Lee, Brooklyn, NY (US); and John B. Putman, Celebration, FL (US)
Assigned to Nanotronics Imaging, Inc., Cuyahoga Falls, OH (US)
Filed by Nanotronics Imaging, Inc., Cuyahoga Falls, OH (US)
Filed on Aug. 14, 2023, as Appl. No. 18/449,320.
Application 18/449,320 is a continuation of application No. 17/938,885, filed on Oct. 7, 2022, granted, now 11,727,672.
Claims priority of provisional application 63/365,247, filed on May 24, 2022.
Prior Publication US 2023/0394801 A1, Dec. 7, 2023
This patent is subject to a terminal disclaimer.
Int. Cl. G06V 10/774 (2022.01); G06T 7/00 (2017.01); G06V 10/22 (2022.01)
CPC G06V 10/774 (2022.01) [G06T 7/001 (2013.01); G06V 10/235 (2022.01); G06T 2207/20104 (2013.01)] 17 Claims
OG exemplary drawing
 
1. A method, comprising:
causing, by a computing system, presentation of graphical user interface comprising an image of a specimen to a user, the image comprising a defect on the specimen;
receiving, by the computing system, a user input annotating the image on the graphical user interface, wherein the user input comprises a first indication of a location of the defect on the specimen and a second indication of a class corresponding to the defect;
automatically adjusting, by the computing system, the first indication of the location of the defect on the specimen in accordance with an error profile corresponding to the user, the error profile generated by training a machine learning model to learn an individualized defect label pattern associated with the user based on a plurality of training annotated images of specimens associated with the user, the error profile defining adjustments to annotations, including the user input, generated by the user based on a determined class corresponding to the defect; and
updating, by the computing system, the graphical user interface to display of the adjusted first indication to the user.