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

深層学習を用いた弱教師あり学習による画像に対する物体位置推定

Research Project

Project/Area Number 17J10261
Research Category

Grant-in-Aid for JSPS Fellows

Allocation TypeSingle-year Grants
Section国内
Research Field Perceptual information processing
Research InstitutionThe University of Electro-Communications

Principal Investigator

下田 和  電気通信大学, 情報理工学研究科, 特別研究員(DC1)

Project Period (FY) 2017-04-26 – 2020-03-31
Project Status Completed (Fiscal Year 2019)
Budget Amount *help
¥2,500,000 (Direct Cost: ¥2,500,000)
Fiscal Year 2019: ¥800,000 (Direct Cost: ¥800,000)
Fiscal Year 2018: ¥800,000 (Direct Cost: ¥800,000)
Fiscal Year 2017: ¥900,000 (Direct Cost: ¥900,000)
Keywords弱教師あり領域分割 / 領域分割 / 深層学習 / 弱教師あり学習
Outline of Annual Research Achievements

本研究課題は深層学習を用いた弱教師あり学習による画像に対する物体位置推定である。深層学習における教師情報のコスト削減についての研究は近年注目を集めているが、本研究課題は特に領域分割における教師情報の削減方法についての研究を行っている。領域分割はComputer Visionにおいて長く研究されてきた重要なテーマであり、他のCV分野のタスクと比較して深層学習における教師情報のコストが高くこれのコスト削減が可能となれば大きな利益になると期待できる。
本研究ではこれを達成するために、弱教師あり領域分割に着目をおき研究を行った。弱教師あり領域分割はクラスラベルから領域分割を学習するアプローチである。クラスラベルは画像に映っている対象物体のタグ情報のことである。これはピクセルレベルのアノテーションと比較してコストが安価であるため、これを用いて領域分割モデルが学習できれば大きな学習コスト削減となる。本研究においては弱教師あり領域分割の問題を教師情報となる擬似領域分割ラベルの精度向上、ひいては擬似領域分割ラベルのノイズ除去問題であると置き換えて考えた。そして、以下の二つのアプローチについて検証を行った。
(1)領域分割の容易性の推定値を使った画像レベルのノイズ除去による精度向上
(2)自己教師あり学習による変化領域の推論を活用した領域レベルのノイズ除去による精度向上
特に、自己教師あり学習による変化領域の推論を活用した領域レベルのノイズ除去による精度向上の研究においては、現在の弱教師あり領域分割のベンチマークにおける最高精度を達成した。この研究成果はComputer Vision分野におけるトップカンファレンスであるICCV 2019(Acceptance rate 25%)に採択され、弱教師あり領域分割の進歩に貢献したと判断できる。

Research Progress Status

令和元年度が最終年度であるため、記入しない。

Strategy for Future Research Activity

令和元年度が最終年度であるため、記入しない。

Report

(3 results)
  • 2019 Annual Research Report
  • 2018 Annual Research Report
  • 2017 Annual Research Report
  • Research Products

    (26 results)

All 2020 2019 2018 2017

All Journal Article (2 results) (of which Peer Reviewed: 2 results,  Open Access: 1 results) Presentation (24 results) (of which Int'l Joint Research: 16 results)

  • [Journal Article] Webly-Supervised Food Detection with Foodness Proposal2019

    • Author(s)
      Wataru Shimoda and Keiji Yanai
    • Journal Title

      IEICE Transactions on Information and Systems

      Volume: E102.D Issue: 7 Pages: 1230-1239

    • DOI

      10.1587/transinf.2018CEP0001

    • NAID

      130007671323

    • ISSN
      0916-8532, 1745-1361
    • Year and Date
      2019-07-01
    • Related Report
      2018 Annual Research Report
    • Peer Reviewed / Open Access
  • [Journal Article] Weakly Supervised Semantic Segmentation Using Distinct Class Specific Saliency Maps2019

    • Author(s)
      Wataru Shimoda and Keiji Yanai
    • Journal Title

      Computer Vision and Image Understanding

      Volume: - Pages: 102712-102712

    • DOI

      10.1016/j.cviu.2018.08.006

    • Related Report
      2018 Annual Research Report
    • Peer Reviewed
  • [Presentation] Predicting Plate Regions for Weakly-supervised Food Image Segmentation2020

    • Author(s)
      Wataru Shimoda and Keiji Yanai
    • Organizer
      Proc. of IEEE International Conference on Multimedia and Expo (ICME)
    • Related Report
      2019 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Self-supervised Difference Detection for Refinement CRF and Seed Interpolation2019

    • Author(s)
      Wataru Shimoda and Keiji Yanai
    • Organizer
      Proc. of CVPR Workshop on Weakly Supervised Learning for Real-World Computer Vision Applications
    • Related Report
      2019 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Unseen Food Creation by Mixing Existing Food Images with Conditional StyleGAN2019

    • Author(s)
      Daichi Horita, Wataru Shimoda and Keiji Yanai
    • Organizer
      Proc. of ACMMM Workshop on Multimedia Assisted Dietary Management (MADIMA)
    • Related Report
      2019 Annual Research Report
    • Int'l Joint Research
  • [Presentation] A New Large-scale Food Image Segmentation Dataset and Its Application to Food Calorie Estimation Based on Grains of Rice2019

    • Author(s)
      Takumi Ege, Wataru Shimoda and Keiji Yanai
    • Organizer
      Proc. of ACMMM Workshop on Multimedia Assisted Dietary Management (MADIMA)
    • Related Report
      2019 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Ramen as You Like: Sketch-based Food Image Generation and Editing2019

    • Author(s)
      Jaehyeong Cho, Wataru Shimoda and Keiji Yanai
    • Organizer
      Proc. of ACM Multimedia demo paper
    • Related Report
      2019 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Zero-Annotation Plate Segmentation Using a Food Category Classifier and a Food/Non-Food Classifier2019

    • Author(s)
      Wataru Shimoda and Keiji Yanai
    • Organizer
      Proc. of ICCV Workshop on Multi-Discipline Approarch for Learning Concepts (MDALC)
    • Related Report
      2019 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Self-Supervised Difference Detection for Weakly-Supervised Semantic Segmentation2019

    • Author(s)
      Wataru Shimoda and Keiji Yanai
    • Organizer
      Proc. of IEEE/CVF Intternational Conference on Computer Vision (ICCV)
    • Related Report
      2019 Annual Research Report
    • Int'l Joint Research
  • [Presentation] A Large-scale Analysis of Regional Tendency of Twitter Photos Using Only Image Features2019

    • Author(s)
      Tetsuya Nagano, Takumi Ege, Wataru Shimoda and Keiji Yanai
    • Organizer
      Proc. of IEEE International Conference on Multimedia Information Processing and Retrieval (MIPR)
    • Related Report
      2018 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Image-Based Estimation of Real Food Size for Accurate Food Calorie Estimation2019

    • Author(s)
      Takumi Ege, Yoshikazu Ando, Ryosuke Tanno, Wataru Shimoda and Keiji Yanai
    • Organizer
      Proc. of IEEE International Conference on Multimedia Information Processing and Retrieval (MIPR)
    • Related Report
      2018 Annual Research Report
    • Int'l Joint Research
  • [Presentation] 変化領域の推測による弱教師あり領域分割の精度向上2019

    • Author(s)
      下田 和, 柳井 啓司
    • Organizer
      電子情報通信学会 パターン認識・メディア理解研究会(PRMU)
    • Related Report
      2018 Annual Research Report
  • [Presentation] Food Category Transfer with Conditional Cycle GAN and a Large-scale Food Image Dataset2018

    • Author(s)
      Daichi Horita, Ryosuke Tanno, Wataru Shimoda, Keiji Yanai
    • Organizer
      Proc. of International Workshop on Multimedia Assisted Dietary Management (MADIMA)
    • Related Report
      2018 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Food Image Generation using A Large Amount of Food Images with Conditional GAN: RamenGAN and RecipeGAN2018

    • Author(s)
      Yoshifumi Ito, Wataru Shimoda and Keiji Yanai
    • Organizer
      Proc. of International Workshop on Multimedia Assisted Dietary Management (MADIMA)
    • Related Report
      2018 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Magical Rice Bowl: Real-time Food Category Changer2018

    • Author(s)
      Ryosuke Tanno, Daichi Horita, Wataru Shimoda and Keiji Yanai
    • Organizer
      ACM Multimedia, Demo Paper
    • Related Report
      2018 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Font Style Transfer Using Neural Style Transfer and Unsupervised Cross-domain Transfer2018

    • Author(s)
      Atsushi Narusawa, Wataru Shimoda, and Keiji Yanai
    • Organizer
      Proc. of ACCV Workshop on AI Aesthetics in Art and Media
    • Related Report
      2018 Annual Research Report
    • Int'l Joint Research
  • [Presentation] 深層学習による質感文字生成2018

    • Author(s)
      成沢 淳史, 下田 和, 柳井 啓司
    • Organizer
      人工知能学会全国大会
    • Related Report
      2018 Annual Research Report
  • [Presentation] 大量のTwitter画像を用いたConditional Cycle GANによる食事写真カテゴリ変換2018

    • Author(s)
      堀田 大地, 成冨 志優, 丹野 良介, 下田 和, 柳井 啓司
    • Organizer
      人工知能学会全国大会
    • Related Report
      2018 Annual Research Report
  • [Presentation] 教師情報に含まれるノイズに堅牢な弱教師あり領域分割手法2018

    • Author(s)
      下田 和, 柳井 啓司
    • Organizer
      画像の認識・理解シンポジウム(MIRU)
    • Related Report
      2018 Annual Research Report
  • [Presentation] 画像マイニングを用いた Conditional Cycle GAN による食事画像変換2018

    • Author(s)
      堀田 大地, 丹野 良介, 下田 和, 柳井 啓司
    • Organizer
      画像の認識・理解シンポジウム(MIRU)
    • Related Report
      2018 Annual Research Report
  • [Presentation] CNNを用いた質感文字生成2018

    • Author(s)
      成沢淳史, 下田和, 柳井啓司
    • Organizer
      画像の認識・理解シンポジウム(MIRU)
    • Related Report
      2018 Annual Research Report
  • [Presentation] Predicting Segmentation ``Easiness'' from the Consistency for Weakly-Supervised Segmentation2017

    • Author(s)
      Wataru Shimoda, Keiji Yanai
    • Organizer
      Proc. of Asian Conference on Pattern Recognition (ACPR)
    • Related Report
      2017 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Learning Food Image Similarity for Food Image Retrieval2017

    • Author(s)
      Wataru Shimoda, Keiji Yanai
    • Organizer
      International Conference on Multimedia Big Data (BIGMM)
    • Related Report
      2017 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Partial Style Transfer Using Weakly-Supervised Semantic Segmentation2017

    • Author(s)
      Shin Matsuo, Wataru Shimoda, Keiji Yanai
    • Organizer
      ICME Workshop on Multimedia Artworks Analysis (MMArt)
    • Related Report
      2017 Annual Research Report
    • Int'l Joint Research
  • [Presentation] 弱教師あり領域分割のための一貫性に基づく学習画像の領域分割容易性推定2017

    • Author(s)
      下田 和, 柳井 啓司
    • Organizer
      画像の認識・理解シンポジウム(MIRU)
    • Related Report
      2017 Annual Research Report
  • [Presentation] 完全教師あり学習手法を用いた弱教師あり領域分割におけるシード領域生成方法の改良2017

    • Author(s)
      下田 和, 柳井 啓司
    • Organizer
      情報処理学会 コンピュータビジョンとイメージメディア研究会 (CVIM)
    • Related Report
      2017 Annual Research Report

URL: 

Published: 2017-05-25   Modified: 2024-03-26  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi