US 12,080,043 B2
Enhanced image object detection using semantic knowledge graph augmentation
Freddy Lecue, Montreal (CA); David Beach, Toronto (CA); and Tanguy Pommellet, Toronto (CA)
Assigned to THALES CANADA INC., Toronto (CA)
Filed by THALES CANADA INC., Toronto (CA)
Filed on Feb. 24, 2021, as Appl. No. 17/184,159.
Claims priority of provisional application 62/980,657, filed on Feb. 24, 2020.
Prior Publication US 2021/0264226 A1, Aug. 26, 2021
Int. Cl. G06V 10/44 (2022.01); G06F 16/901 (2019.01); G06N 3/042 (2023.01); G06N 3/08 (2023.01); G06V 10/82 (2022.01); G06V 30/19 (2022.01); G06V 30/262 (2022.01); G06V 20/58 (2022.01)
CPC G06V 10/454 (2022.01) [G06F 16/9024 (2019.01); G06N 3/042 (2023.01); G06N 3/08 (2013.01); G06V 10/82 (2022.01); G06V 30/19173 (2022.01); G06V 30/274 (2022.01); G06V 20/58 (2022.01)] 20 Claims
OG exemplary drawing
 
1. A method of augmented semantic object detection in an image dataset comprising:
receiving an object detection task including expected labels;
selecting a knowledge graph that includes the expected labels;
extracting semantic links relevant to the image dataset with the knowledge graph;
detecting objects in the image dataset using a regional proposal network;
determining initial predictions of detected objects, where the initial predictions of detected objects have confidence scores;
comparing the confidence scores of the initial predictions of detected objects with a threshold;
selecting detected objects from the initial predictions of detected objects for augmentation;
comparing the selected detected objects with the semantic links; and
augmenting the confidence scores based on the semantic links between the detected objects.