Unconstrained and Calibration-free Gaze Estimation in a Room-scale Area using a Monocular Camera

Kimimasa Tamura, Ran Choi, Yoshimitsu Aoki

研究成果: Article

2 引用 (Scopus)

抄録

Gaze estimation using monocular cameras has high industrial application value and many studies have been undertaken on head pose-invariant and calibration-free gaze estimation. Head positions in existing datasets used in these studies are, however, limited to the vicinity of the camera and methods trained on such datasets are not applicable when subjects are distant from the camera. In this study, we create a room-scale gaze dataset with largely varied head poses to achieve a robust gaze estimation in broader range of width and depths. Head positions are much farther and the resolution of eye image is smaller than that in conventional datasets. To address this issue, we propose a likelihood evaluation method based on edge gradient with dense particles for iris tracking to achieve robust tracking at low resolution eye image. Our proposed method has been proven more accurate than conventional methods on all the individuals in our dataset through several experiments with cross-validation.

元の言語English
ジャーナルIEEE Access
DOI
出版物ステータスAccepted/In press - 2017 8 5

Fingerprint

Cameras
Calibration
Industrial applications
Experiments

ASJC Scopus subject areas

  • Computer Science(all)
  • Materials Science(all)
  • Engineering(all)

これを引用

@article{986c18ed5e5d4f709c35ae4732b813e6,
title = "Unconstrained and Calibration-free Gaze Estimation in a Room-scale Area using a Monocular Camera",
abstract = "Gaze estimation using monocular cameras has high industrial application value and many studies have been undertaken on head pose-invariant and calibration-free gaze estimation. Head positions in existing datasets used in these studies are, however, limited to the vicinity of the camera and methods trained on such datasets are not applicable when subjects are distant from the camera. In this study, we create a room-scale gaze dataset with largely varied head poses to achieve a robust gaze estimation in broader range of width and depths. Head positions are much farther and the resolution of eye image is smaller than that in conventional datasets. To address this issue, we propose a likelihood evaluation method based on edge gradient with dense particles for iris tracking to achieve robust tracking at low resolution eye image. Our proposed method has been proven more accurate than conventional methods on all the individuals in our dataset through several experiments with cross-validation.",
keywords = "Gaze Estimation, Iris Tracking, Particle Filter, Regression",
author = "Kimimasa Tamura and Ran Choi and Yoshimitsu Aoki",
year = "2017",
month = "8",
day = "5",
doi = "10.1109/ACCESS.2017.2734168",
language = "English",
journal = "IEEE Access",
issn = "2169-3536",
publisher = "Institute of Electrical and Electronics Engineers Inc.",

}

TY - JOUR

T1 - Unconstrained and Calibration-free Gaze Estimation in a Room-scale Area using a Monocular Camera

AU - Tamura, Kimimasa

AU - Choi, Ran

AU - Aoki, Yoshimitsu

PY - 2017/8/5

Y1 - 2017/8/5

N2 - Gaze estimation using monocular cameras has high industrial application value and many studies have been undertaken on head pose-invariant and calibration-free gaze estimation. Head positions in existing datasets used in these studies are, however, limited to the vicinity of the camera and methods trained on such datasets are not applicable when subjects are distant from the camera. In this study, we create a room-scale gaze dataset with largely varied head poses to achieve a robust gaze estimation in broader range of width and depths. Head positions are much farther and the resolution of eye image is smaller than that in conventional datasets. To address this issue, we propose a likelihood evaluation method based on edge gradient with dense particles for iris tracking to achieve robust tracking at low resolution eye image. Our proposed method has been proven more accurate than conventional methods on all the individuals in our dataset through several experiments with cross-validation.

AB - Gaze estimation using monocular cameras has high industrial application value and many studies have been undertaken on head pose-invariant and calibration-free gaze estimation. Head positions in existing datasets used in these studies are, however, limited to the vicinity of the camera and methods trained on such datasets are not applicable when subjects are distant from the camera. In this study, we create a room-scale gaze dataset with largely varied head poses to achieve a robust gaze estimation in broader range of width and depths. Head positions are much farther and the resolution of eye image is smaller than that in conventional datasets. To address this issue, we propose a likelihood evaluation method based on edge gradient with dense particles for iris tracking to achieve robust tracking at low resolution eye image. Our proposed method has been proven more accurate than conventional methods on all the individuals in our dataset through several experiments with cross-validation.

KW - Gaze Estimation

KW - Iris Tracking

KW - Particle Filter

KW - Regression

UR - http://www.scopus.com/inward/record.url?scp=85028927970&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85028927970&partnerID=8YFLogxK

U2 - 10.1109/ACCESS.2017.2734168

DO - 10.1109/ACCESS.2017.2734168

M3 - Article

AN - SCOPUS:85028927970

JO - IEEE Access

JF - IEEE Access

SN - 2169-3536

ER -