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

Gaze Estimator Domain Adaptation Based on Data Generation Through Face Shape Reconstruction and Self-Supervised Auxiliary Tasks

Research Project

Project/Area Number 21K11932
Research Category

Grant-in-Aid for Scientific Research (C)

Allocation TypeMulti-year Fund
Section一般
Review Section Basic Section 61010:Perceptual information processing-related
Research InstitutionThe University of Tokyo

Principal Investigator

Sugano Yusuke  東京大学, 生産技術研究所, 准教授 (10593585)

Project Period (FY) 2021-04-01 – 2024-03-31
Project Status Completed (Fiscal Year 2023)
Budget Amount *help
¥4,160,000 (Direct Cost: ¥3,200,000、Indirect Cost: ¥960,000)
Fiscal Year 2023: ¥1,560,000 (Direct Cost: ¥1,200,000、Indirect Cost: ¥360,000)
Fiscal Year 2022: ¥1,560,000 (Direct Cost: ¥1,200,000、Indirect Cost: ¥360,000)
Fiscal Year 2021: ¥1,040,000 (Direct Cost: ¥800,000、Indirect Cost: ¥240,000)
Keywords視線推定 / 機械学習 / ドメイン適応 / コンピュータビジョン
Outline of Research at the Start

画像入力のみを手がかりに人物の視線方向を推定するための技術は現在活発に研究が行われているが、学習時と推論時の環境の違い、特にカメラから見た頭部姿勢・視線方向の分布の違いにより推定性能が劣化する問題を抱えている。本研究では、既存の視線データセットから復元した顔形状を元に姿勢を変えてレンダリングした学習データを元に、生成画像と実画像の見え方の違いを吸収するためのドメイン適応手法の開発を行うことで、学習データに含まれない頭部姿勢範囲にも対応できる視線推定手法の確立を目指す。

Outline of Final Research Achievements

In this study, we combined a data generation method based on 3D face shape reconstruction with a domain adaptation technique using feature separation to develop a robust gaze estimation model that operates effectively in unknown environments. By reconstructing face shapes from monocular images and rendering them in various orientations, we enhanced the diversity of the training data. Unsupervised domain adaptation was employed to bridge the gap between generated data and real data. Additionally, we developed an appearance-based gaze estimation model using multi-camera input, achieving high generalization performance through feature transformation and fusion based on the relative orientation between cameras.

Academic Significance and Societal Importance of the Research Achievements

本研究の学術的意義は、顔形状の3次元復元とドメイン適応を組み合わせた新しい視線推定手法を提案し、未知の環境でも高い精度を実現したことにある。また、任意の複数カメラを用いて視線推定を行うことのできる手法には前例がなく、カメラの位置関係を拘束条件として用いる特徴融合は他の課題にも応用できる可能性がある。
提案手法により、多様な姿勢や環境での視線推定が可能となり、自然なインタラクションを必要とする様々なアプリケーションに活用できる。例えば、対話システムやデジタルサイネージ、自動車の運転支援など、ユーザの視線情報を用いることで、よりシームレスで直感的なインターフェースの実現が期待できる。

Report

(4 results)
  • 2023 Annual Research Report   Final Research Report ( PDF )
  • 2022 Research-status Report
  • 2021 Research-status Report
  • Research Products

    (11 results)

All 2024 2023 2022 2021

All Journal Article (3 results) (of which Peer Reviewed: 3 results,  Open Access: 3 results) Presentation (7 results) (of which Int'l Joint Research: 3 results,  Invited: 2 results) Patent(Industrial Property Rights) (1 results)

  • [Journal Article] Rotation-Constrained Cross-View Feature Fusion for Multi-View Appearance-based Gaze Estimation2024

    • Author(s)
      Yoichiro Hisadome、Tianyi Wu、Jiawei Qin、Yusuke Sugano
    • Journal Title

      Proc. 2024 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)

      Volume: 1 Pages: 5973-5982

    • DOI

      10.1109/wacv57701.2024.00588

    • Related Report
      2023 Annual Research Report
    • Peer Reviewed / Open Access
  • [Journal Article] Learning Video-Independent Eye Contact Segmentation from?In-the-Wild Videos2023

    • Author(s)
      Wu Tianyi、Sugano Yusuke
    • Journal Title

      Lecture Notes in Computer Science (ACCV2022)

      Volume: 13844 Pages: 52-70

    • DOI

      10.1007/978-3-031-26316-3_4

    • ISBN
      9783031263156, 9783031263163
    • Related Report
      2022 Research-status Report
    • Peer Reviewed / Open Access
  • [Journal Article] Learning-by-Novel-View-Synthesis for Full-Face Appearance-Based 3D Gaze Estimation2022

    • Author(s)
      Qin Jiawei、Shimoyama Takuru、Sugano Yusuke
    • Journal Title

      Proc. 2022 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW)

      Volume: - Pages: 4977-4987

    • DOI

      10.1109/cvprw56347.2022.00546

    • Related Report
      2022 Research-status Report
    • Peer Reviewed / Open Access
  • [Presentation] Rotation-Constrained Cross-View Feature Fusion for Multi-View Appearance-based Gaze Estimation2024

    • Author(s)
      Yoichiro Hisadome
    • Organizer
      2024 IEEE/CVF Winter Conference on Applications of Computer Vision (WACV)
    • Related Report
      2023 Annual Research Report
    • Int'l Joint Research
  • [Presentation] Toward Appearance-Based Gaze Estimation Open To Diverse People And Environments2023

    • Author(s)
      Yusuke Sugano
    • Organizer
      2023 ACM Symposium of Eye Tracking Research & Applications (ETRA)
    • Related Report
      2023 Annual Research Report
    • Invited
  • [Presentation] 人にひらかれたメディア理解に向けて ―人を理解する、人と理解する―2022

    • Author(s)
      菅野裕介
    • 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
      2022 Research-status Report
    • Invited
  • [Presentation] View-consistent Feature Alignment for Multi-view Appearance-based Gaze Estimation2022

    • Author(s)
      Yoichiro Hisadome、Yusuke Sugano
    • Organizer
      第25回 画像の認識・理解シンポジウム
    • Related Report
      2022 Research-status Report
  • [Presentation] Learning-by-Novel-View-Synthesis for Full-Face Appearance-Based 3D Gaze Estimation2022

    • Author(s)
      Qin Jiawei、Shimoyama Takuru、Sugano Yusuke
    • Organizer
      4th International Workshop on Gaze Estimation and Prediction in the Wild
    • Related Report
      2022 Research-status Report
    • Int'l Joint Research
  • [Presentation] Learning Video-Independent Eye Contact Segmentation from?In-the-Wild Videos2022

    • Author(s)
      Wu Tianyi、Sugano Yusuke
    • Organizer
      16th Asian Conference on Computer Vision (ACCV2022)
    • Related Report
      2022 Research-status Report
    • Int'l Joint Research
  • [Presentation] ラベル分布の異なるドメインに対するアピアランスベース視線推定モデルの教師無し適応2021

    • Author(s)
      下山拓流、菅野裕介
    • 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
      2021 Research-status Report
  • [Patent(Industrial Property Rights)] 学習モデル生成プログラム、情報処理装置及び学習 モデル生成方法2023

    • Inventor(s)
      菅野 裕介、久留 陽一郎、呉 天一、秦 嘉偉
    • Industrial Property Rights Holder
      菅野 裕介、久留 陽一郎、呉 天一、秦 嘉偉
    • Industrial Property Rights Type
      特許
    • Filing Date
      2023
    • Related Report
      2023 Annual Research Report

URL: 

Published: 2021-04-28   Modified: 2025-01-30  

Information User Guide FAQ News Terms of Use Attribution of KAKENHI

Powered by NII kakenhi