US 12,271,442 B2
Semantic image segmentation for cognitive analysis of physical structures
Krishna Dev Oruganty, Long Island, NY (US); Siva Tian, New York, NY (US); Amit Arora, Princeton, NJ (US); and Edmond John Schneider, Henrico, VA (US)
Assigned to Genpact USA, Inc., New York, NY (US)
Filed by Genpact USA, Inc, New York, NY (US)
Filed on Jul. 30, 2021, as Appl. No. 17/390,303.
Claims priority of provisional application 63/059,482, filed on Jul. 31, 2020.
Prior Publication US 2022/0036132 A1, Feb. 3, 2022
Int. Cl. G06F 18/214 (2023.01); G06F 18/20 (2023.01); G06F 18/21 (2023.01); G06F 18/24 (2023.01); G06T 7/00 (2017.01); G06N 3/04 (2023.01)
CPC G06F 18/214 (2023.01) [G06F 18/2163 (2023.01); G06F 18/217 (2023.01); G06F 18/24 (2023.01); G06F 18/285 (2023.01); G06T 7/0002 (2013.01); G06N 3/04 (2013.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01); G06V 2201/08 (2022.01)] 32 Claims
OG exemplary drawing
 
1. A computer-implemented method, comprising:
selecting an artificial intelligence or machine learning (AI/ML) system of a particular type customized for identifying parts of a particular orientation or a particular structure type;
configuring the selected AI/ML system using a parts-identification model stored for a previously-trained AI/ML system of the particular type;
performing auto-labeling to generate one or more auto-labeled images using the selected AI/ML system based at least on one of the orientation or structure type;
generating a confidence score for labeling the one or more auto-labeled images based on one of a size constraint or a geometry constraint;
training the AI/ML system using a dataset comprising the one or more auto-labeled images having a respective confidence score exceeding a specified threshold; and
outputting and storing configuration of the trained and selected AI/ML system as an improved parts-identification model.