公开(公告)号：US20220327675A1

公开(公告)日：2022-10-13

申请号：US17595608

申请日：2020-04-24

Applicant: SHOWA DENKO K.K.

Inventor： Yoshishige OKUNO ,  Eriko TAKEDA ,  Junichi SHIBUYA ,  Yuki NAKAO ,  Kenji OGAWA

IPC: G06T7/00 ,  G06T7/60 ,  G06V10/74 ,  G06V10/75 ,  G01N15/14

Abstract: To improve a determination accuracy when determining each particle contained in an image of an object. An image analysis apparatus according to an embodiment of the present invention includes: a shape determination unit configured to determine a shape of a particle included in a particle image that is extracted from an image of an object, so that an OK particle image which is a particle image of an OK particle that satisfies a predetermined standard for shape and a provisional NG particle image which is a particle image of a provisional NG particle that does not satisfy the predetermined standard, are obtained; a pseudo image generation unit configured to generate a pseudo image; and a similarity determination unit configured to determine whether the provisional NG image and the pseudo image are similar.

公开(公告)号：US20220391759A1

公开(公告)日：2022-12-08

申请号：US17756002

申请日：2020-11-12

Applicant: SHOWA DENKO K.K.

Inventor： Yoshishige OKUNO ,  Eriko TAKEDA ,  Junichi SHIBUYA ,  Yuki NAKAO ,  Kenji OGAWA

IPC: G06N20/00

Abstract: The present invention is designed to improve the accuracy of machine learning. The present invention provides: a first classification unit configured to classify data into classifiable data and initially-unclassifiable data based on a first learning model; a first annotation unit configured to annotate the classifiable data with a label; a second classification unit configured to classify the initially-unclassifiable data based on a second learning model; a label acquiring unit configured to acquire a label with which the initially-unclassifiable data is to be annotated; a second annotation unit configured to annotate the initially-unclassifiable data with a label; and a second learning model updating unit configured to update the second learning model based on the initially-unclassifiable data that is annotated with the label based on a result of classification by the second classification unit and the label acquired by the label acquiring unit.