• Search Research Projects
  • Search Researchers
  • How to Use
  1. Back to project page

2022 Fiscal Year Final Research Report

Research on construction of highly accurate image recognition methods from limited supervised data

Research Project

  • PDF
Project/Area Number 19H01115
Research Category

Grant-in-Aid for Scientific Research (A)

Allocation TypeSingle-year Grants
Section一般
Review Section Medium-sized Section 61:Human informatics and related fields
Research InstitutionThe University of Tokyo

Principal Investigator

Harada Tatsuya  東京大学, 大学院情報理工学系研究科, 教授 (60345113)

Project Period (FY) 2019-04-01 – 2023-03-31
Keywords画像認識 / 機械学習
Outline of Final Research Achievements

Recent successes in deep learning have dramatically improved the accuracy of image recognition but achieving high recognition performance requires a huge amount of supervised data. Generating high-quality supervised data requires a lot of human effort and cost, which is a major problem in machine learning. In this study, we developed a method for learning highly accurate image recognition models with only a small amount of supervised data. Specifically, we developed a methodology to maximize the discriminative power of deep learning by making the most of limited supervised data, a domain adaptation method that enables knowledge transfer between different domains, and active information acquisition for efficient generation of supervised data.

Free Research Field

知能機械情報学

Academic Significance and Societal Importance of the Research Achievements

現在成功している高精度の画像認識システムは教師あり学習を基盤としているが,大変なコストがかかるため機械学習分野において大問題となっている.さらに,付与するラベルに高度な専門知識を必要とする場合,アノテーションができる人が少数であり,膨大な教師データを作ることが不可能に近い.以上のように,教師データが入手困難な状況は多方面で存在する.従って,少数の教師データから高精度な画像認識モデルを学習するための方法論の実現は,現状の知的なシステムがより汎用的に利用されるための学術的,社会的最重要課題の一つであり,本研究成果はこの問題解決の一翼を担うものである.

URL: 

Published: 2024-01-30  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi