US 11,657,491 B2
Learning data collection apparatus, learning data collection method, and program
Shuhei Horita, Tokyo (JP)
Assigned to FUJIFILM Corporation, Tokyo (JP)
Filed by FUJIFILM Corporation, Tokyo (JP)
Filed on Mar. 16, 2021, as Appl. No. 17/203,198.
Application 17/203,198 is a continuation of application No. PCT/JP2019/036369, filed on Sep. 17, 2019.
Claims priority of application No. JP2018-176316 (JP), filed on Sep. 20, 2018.
Prior Publication US 2021/0209422 A1, Jul. 8, 2021
Int. Cl. G06K 9/00 (2022.01); G06T 7/00 (2017.01); G06F 18/21 (2023.01); G06V 10/25 (2022.01); G06V 10/771 (2022.01); G06V 10/774 (2022.01); G06V 10/778 (2022.01); G06V 10/50 (2022.01); G06V 10/44 (2022.01); G06V 10/46 (2022.01)
CPC G06T 7/0004 (2013.01) [G06F 18/2178 (2023.01); G06T 7/0002 (2013.01); G06V 10/25 (2022.01); G06V 10/44 (2022.01); G06V 10/50 (2022.01); G06V 10/771 (2022.01); G06V 10/7747 (2022.01); G06V 10/7784 (2022.01); G06T 2207/30168 (2013.01); G06V 10/467 (2022.01)] 12 Claims
OG exemplary drawing
1. A learning data collection apparatus comprising
at least one processor configured to:
acquire an inspection image, the inspection image being a captured image of an object to be inspected;
acquire a region detection result on the basis of the inspection image, the region detection result indicating a region detected by a region detector that is trained;
acquire a correction history of the region detection result;
calculate correction quantification information obtained by quantifying the correction history;
store the inspection image, the region detection result, and the correction history in association with each other;
set a threshold value of the correction quantification information as an extraction condition, the extraction condition being a condition for extracting the inspection image to be used for retraining from the database; and
extract, as learning data for retraining the region detector, the inspection image satisfying the extraction condition and the region detection result and the correction history that are associated with the inspection image from the database.