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

Generative adversarial networks for unnatural human motion detection and generation

Research Project

Project/Area Number 19K20310
Research Category

Grant-in-Aid for Early-Career Scientists

Allocation TypeMulti-year Fund
Review Section Basic Section 61010:Perceptual information processing-related
Research InstitutionTokyo Metropolitan Industrial Technology Research Institute

Principal Investigator

Daisuke Miki  地方独立行政法人東京都立産業技術研究センター, 開発本部開発第三部情報技術グループ, 副主任研究員 (70757343)

Project Period (FY) 2019-04-01 – 2021-03-31
Project Status Completed (Fiscal Year 2020)
Budget Amount *help
¥2,470,000 (Direct Cost: ¥1,900,000、Indirect Cost: ¥570,000)
Fiscal Year 2020: ¥910,000 (Direct Cost: ¥700,000、Indirect Cost: ¥210,000)
Fiscal Year 2019: ¥1,560,000 (Direct Cost: ¥1,200,000、Indirect Cost: ¥360,000)
Keywords人物動作解析 / 異常検知 / 深層学習 / 敵対的生成ネットワーク / 監視カメラ
Outline of Research at the Start

監視カメラの設置台数は年々増加し、膨大な映像が蓄積されている。一方で、映像を自動で解析し、人物の不審な動作を検知するような技術は実用化に向けて研究の余地がある。本研究は、監視カメラ映像から不審動作を検出ための学習手法を確立する。正常動作および不審動作の特徴量をそれぞれ反転する反転生成器と、実データおよび生成データを区別する識別器、さらに不審動作の検出器の三者を敵対的に学習させることで、データセットを拡張しながら検出器の学習を行い、不審動作の検出精度と汎化性能の向上を図る。さらに本技術を応用し、データセットに含まれない未知の不審動作を検出する技術の確立を目指す。

Outline of Final Research Achievements

A deep neural network (DNN) model and a training method for analyzing spatial-temporal human motion data are proposed in this study. We confirmed that the proposed method can detect anomalies in human motion data, and the DNN model trained using our multi-instance learning-based training method can detect ambiguously defined motions, such as unnatural human motions. The results indicate that it is possible to detect anomalies and identify human motion when dealing with large datasets that are difficult to annotate locally or when the data contains ambiguous motions, such as dangerous or unnatural motion. To improve the generalization performance of the anomaly detector in the case of a limited number of datasets, we establish a training data generation method using generative adversarial networks.

Academic Significance and Societal Importance of the Research Achievements

申請者らはこれまでに監視カメラ映像から人物の挙動や軌跡等の動作特徴量を取得するための画像処理技術の開発に従事してきた。監視カメラの設置台数は年々増加し、膨大な数の映像が蓄積されていることから、これらの映像に含まれる人物の動作特徴量から異常を検知する手法を確立できれば、映像監視の自動化が期待される。本研究では、人物動作特徴量のような多変量時系列データから異常を検出するための、DNNモデルおよびその学習手法の確立を行い、人物動作データからの異常検知を可能とした。さらに数に限りのあるデータセットを用いて異常検出器の汎化性能を改善するための、学習データの拡張手法を確立した。

Report

(3 results)
  • 2020 Annual Research Report   Final Research Report ( PDF )
  • 2019 Research-status Report
  • Research Products

    (9 results)

All 2020 2019

All Journal Article (3 results) (of which Peer Reviewed: 2 results) Presentation (5 results) (of which Int'l Joint Research: 5 results) Patent(Industrial Property Rights) (1 results)

  • [Journal Article] Robust human motion recognition from wide-angle images for video surveillance in nuclear power plants2020

    • Author(s)
      Daisuke Miki, Shinya Abe, Shi Chen, Kazuyuki Demachi
    • Journal Title

      Mechanical Engineering Journal

      Volume: 7 Issue: 3 Pages: 19-00533-19-00533

    • DOI

      10.1299/mej.19-00533

    • NAID

      130007855159

    • ISSN
      2187-9745
    • Related Report
      2020 Annual Research Report 2019 Research-status Report
    • Peer Reviewed
  • [Journal Article] Bearing fault diagnosis using weakly supervised long short-term memory2020

    • Author(s)
      Daisuke Miki, Kazuyuki Demachi
    • Journal Title

      Journal of Nuclear Science and Technology

      Volume: 57 Issue: 9 Pages: 1091-1100

    • DOI

      10.1080/00223131.2020.1761473

    • Related Report
      2020 Annual Research Report 2019 Research-status Report
  • [Journal Article] Robust human pose estimation from distorted wide-angle images through iterative search of transformation parameters2019

    • Author(s)
      Miki Daisuke、Abe Shinya、Chen Shi、Demachi Kazuyuki
    • Journal Title

      Signal, Image and Video Processing

      Volume: 14 Issue: 4 Pages: 693-700

    • DOI

      10.1007/s11760-019-01602-5

    • Related Report
      2019 Research-status Report
    • Peer Reviewed
  • [Presentation] Unnatural Human Motion Detection using Weakly Supervised Deep Neural Network2020

    • Author(s)
      Daisuke Miki, Shi Chen, Kazuyuki Demachi
    • Organizer
      2020 Third International Conference on Artificial Intelligence for Industries (AI4I)
    • Related Report
      2020 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Weakly Supervised Deep Neural Network for Bearing Fault Diagnosis2020

    • Author(s)
      Daisuke Miki, Kazuyuki Demachi
    • Organizer
      The Proceedings of the 2020 International Conference on Nuclear Engineering (ICONE)
    • Related Report
      2020 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Weakly Supervised Graph Convolutional Neural Network for Human Action Localization2020

    • Author(s)
      Daisuke Miki Shi Chen, Kazuyuki Demachi
    • Organizer
      2020 IEEE Winter Conference on Applications of Computer Vision (WACV)
    • Related Report
      2020 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Weakly Supervised Graph Convolutional Neural Network for Human Action Localization2020

    • Author(s)
      Miki Daisuke、Chen Shi、Demachi Kazuyuki
    • Organizer
      2020 IEEE Winter Conference on Applications of Computer Vision (WACV)
    • Related Report
      2019 Research-status Report
    • Int'l Joint Research
  • [Presentation] ROBUST HUMAN MOTION RECOGNITION FROM DISTORTED WIDE-ANGLE IMAGES FOR VIDEO SURVEILLANCE2019

    • Author(s)
      Miki Daisuke、Abe Shinya、Chen Shi、Demachi Kazuyuki
    • Organizer
      27th International Conference on Nuclear Engineering (ICONE)
    • Related Report
      2019 Research-status Report
    • Int'l Joint Research
  • [Patent(Industrial Property Rights)] 行動識別モデル学習装置、行動識別モデル学習方法、行動識別モデル学習プログラム、及び記憶媒体2020

    • Inventor(s)
      三木大輔
    • Industrial Property Rights Holder
      三木大輔
    • Industrial Property Rights Type
      特許
    • Industrial Property Number
      2020-144857
    • Filing Date
      2020
    • Related Report
      2020 Annual Research Report

URL: 

Published: 2019-04-18   Modified: 2022-01-27  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi