LEARNING DEVICE AND LEARNING METHOD

WIPO Patent Application WO/2020/022027

A1

The purpose of the present invention is to provide a learning device and a learning method which enable suitably learning data that belongs to the same category and that was acquired under different conditions. In this learning device, by inputting first and second data into an independent first and second input layer, respectively, for feature value calculation, feature value calculation in one of the first and second input layers is not affected by feature value calculation in the other of the input layers. Further, in addition to feature extraction in the input layers, first intermediate feature value calculation processing and second intermediate feature value calculation processing are performed at least once each in an intermediate layer common to the first and second input layers, whereby the feature values calculated from the first and second data in the input layer can be reflected in intermediate feature value calculation in the intermediate layer. In this way, it is possible to suitably learn data that belongs to the same category and that was acquired under different conditions.

OOSAKE MASAAKI (JP)
OZEKI MAKOTO (JP)

PCT/JP2019/026688

January 30, 2020

July 04, 2019

Click for automatic bibliography generation  

FUJIFILM CORP (JP)

G06N3/08; A61B1/00; A61B1/045; G06N3/04; G06T7/00; G06V10/143; G06V10/32; G06V10/60; G06V10/764; G06V10/776

WO2017158575A1

2017-09-21

JP2016067780A

2016-05-09

ITO, HAYATO ET AL.: "Classification of neoplasia and non-neoplasia for colon endocytoscopic images by convolutional neural network", IEICE TECHNICAL REPORT, vol. 117, no. 220, 18 September 2017 (2017-09-18), pages 17 - 21

MATSUURA, Kenzo (JP)

View/Download PDF      PDF Help

Previous Patent: LENS BARREL, IMAGING DEVICE, POSITION DETECTION METHOD, AND POSITION DETECTION PROGRAM

Next Patent: ACTUATOR