首页> 外国专利> LEARNING PROGRAM, IMAGE CLASSIFICATION PROGRAM, LEARNING METHOD, IMAGE CLASSIFICATION METHOD, LEARNING DEVICE, AND IMAGE CLASSIFICATION DEVICE

LEARNING PROGRAM, IMAGE CLASSIFICATION PROGRAM, LEARNING METHOD, IMAGE CLASSIFICATION METHOD, LEARNING DEVICE, AND IMAGE CLASSIFICATION DEVICE

机译:学习程序,图像分类程序,学习方法,图像分类方法,学习设备和图像分类设备

摘要

To conduct accurate classification cause estimation efficiently.SOLUTION: A learning program according to an embodiment makes a computer execute a process for input, a process of classification, and a process for creation. The process for input inputs each image of a teacher image group including an image group in which a prescribed feature part that is a feature of the classification is included and an image group in which a feature part is not included to a neural network classifying an image. The process for classification classifies an intermediate output at an intermediate layer of the neural network to which each image is inputted. The process for creation creates information indicating a causal relationship from the prescribed feature to a result of the classification depending on which class the intermediate output of the neural network is based on the classification result of the intermediate output of each image.SELECTED DRAWING: Figure 4
机译:为了有效地进行准确的分类原因估计。解决方案:根据实施例的学习程序使计算机执行输入处理,分类处理和创建处理。用于输入的处理将教师图像组的每个图像输入到对图像进行分类的神经网络中,该图像组包括其中包括作为分类的特征的规定特征部分的图像组和不包括特征部分的图像组。 。分类过程对输入每个图像的神经网络中间层的中间输出进行分类。创建过程根据每个图像的中间输出的分类结果,根据神经网络的中间输出是基于哪一类来创建指示从指定特征到分类结果之间因果关系的信息。选择的图形:图4

著录项

  • 公开/公告号JP2020035103A

    专利类型

  • 公开/公告日2020-03-05

    原文格式PDF

  • 申请/专利权人 FUJITSU LTD;

    申请/专利号JP20180159651

  • 发明设计人 NAKAZATO KATSUHISA;

    申请日2018-08-28

  • 分类号G06N20;G06F16/50;G06F16;

  • 国家 JP

  • 入库时间 2022-08-21 11:35:08

相似文献

  • 专利
  • 外文文献
  • 中文文献
获取专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号