首页> 外国专利> AUTONOMOUS LEARNING DEVICE, AUTONOMOUS LEARNING METHOD AND PROGRAM

AUTONOMOUS LEARNING DEVICE, AUTONOMOUS LEARNING METHOD AND PROGRAM

机译:自主学习装置,自主学习方法及程序

摘要

To remove information that becomes noise for object recognition from learning data when generating learning data used for object recognition from information acquired by web search.SOLUTION: In an autonomous learning device 100, a related word group generation unit 1 performs a keyword search using a given input word, extracts a noun from a search result to generate a related word group. A true related word group generation unit 2 determines relevance between a word included in the related word group and an input word, and generates a true related word group in which a word having no predetermined relevance to the input word is removed from the related word group. An image search unit 3 performs an image search using a word included in the true related word group to acquire an image. A classifier generation unit 4 generates a classifier, with each of words included in the true related word group and an image corresponding to each of the words included in the true related word group as learning data.SELECTED DRAWING: Figure 1
机译:当从通过网络搜索获取的信息中生成用于对象识别的学习数据时,从学习数据中去除成为对象识别的噪声的信息。解决方案:在自主学习装置100中,相关单词组生成单元1使用给定单词组进行关键词搜索。输入单词,从搜索结果中提取名词以生成相关单词组。真实相关单词组生成单元2确定相关单词组中包括的单词与输入单词之间的相关性,并生成真实相关单词组,其中从相关单词组中去除与输入单词没有预定相关性的单词。图像搜索单元3使用包括在真实相关单词组中的单词来执行图像搜索以获取图像。分类器生成单元4生成分类器,该分类器将包含在真实相关词组中的每个单词和对应于包含在真实相关词组中的每个单词的图像作为学习数据。

著录项

  • 公开/公告号JP2019067194A

    专利类型

  • 公开/公告日2019-04-25

    原文格式PDF

  • 申请/专利权人 SOINN HOLDINGS LLC;

    申请/专利号JP20170192758

  • 发明设计人 TANAHASHI KOKI;HASEGAWA OSAMU;

    申请日2017-10-02

  • 分类号G06F16;

  • 国家 JP

  • 入库时间 2022-08-21 12:22:09

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