INFORMATION PROCESSOR, METHOD FOR CONTROLLING THE SAME, PROGRAM, AND RECORDING MEDIUM

To meet the demand of efficiently storing a necessary amount of learning data to deal with an issue that it takes much time and labor to prepare learning data for detecting such a deformation as a crack or a defect in an image taken of an inspection target object.SOLUTION: The present invention is m...

Full description

Saved in:
Bibliographic Details
Main Authors MORI KATSUHIKO, HOSHINO SHOICHI, MATSUGI MASAKAZU, NOGAMI ATSUSHI, MITARAI HIROSUKE
Format Patent
LanguageEnglish
Japanese
Published 24.09.2020
Subjects
Online AccessGet full text

Cover

Loading…
More Information
Summary:To meet the demand of efficiently storing a necessary amount of learning data to deal with an issue that it takes much time and labor to prepare learning data for detecting such a deformation as a crack or a defect in an image taken of an inspection target object.SOLUTION: The present invention is made in view of the above objective and includes the steps of: receiving an input of specification of a position of a detection target included in an image (S503); acquiring a storage amount of learning data made of a pair of information indicating the image and information indicating the input position (S504); and displaying, in display means, the storage amount and a reference amount of the learning data in a comparable manner (S505). After the storage amount reaches the reference amount, a learning model for detecting the detection target from the image on the basis of the stored learning data is learned (S509).SELECTED DRAWING: Figure 5 【課題】 検査対象を被写体として撮影した画像からひび割れや欠陥などの変状を検知するための学習データの作成には時間と手間がかかるため、必要な量の学習データを効率的に蓄積したいという要望がある。【解決手段】 本発明は上記課題を鑑みてなされたものであり、画像に含まれる検知対象の位置を指定する入力を受け付け(S503)、前記画像を示す情報と前記入力された位置を示す情報との組からなる学習データの、蓄積量を取得し(S504)、前記蓄積量と前記学習データの基準量とを表示手段に比較可能に表示する(S505)。前記蓄積量が前記基準量に達した後で、蓄積された前記学習データに基づいて前記画像から前記検知対象を検知するための学習モデルを学習する(S509)。【選択図】 図5
Bibliography:Application Number: JP20190055558