Seeing faces in noise: Exploring machine and human face detection processes by the reverse correlation method

Chihiro Saegusa, Megumi Yamaoka, Katsumi Watanabe

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

In the present study, we aimed at investigating possible similarities (and discrepancies) between two major machine algorithms of face detection (AdaBoost and EigenFace) and human face detection processes. For this, we presented the 'false classification images' produced by the two face detection algorithms to human observers. Noise fields were fed into the two algorithms and images in which each algorithm falsely detected faces were collected. Those images were averaged and normalized to obtain false classification images. Human observers performed a psychophysical experiment to detect a face with the false classification images against random noise images. The face detection performance increased almost linearly as the number of averaged false detection images increase. Inverted images reduced the detection performance more with the images produced by EigenFace than those by AdaBoost. The present results suggest that both human and machine detection algorithms tended to make similar errors and therefore both AdaBoost and EigenFace are good approximation of human face processing.

Original languageEnglish
Title of host publication2014 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA 2014
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9786163618238
DOIs
Publication statusPublished - 2014 Feb 12
Externally publishedYes
Event2014 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA 2014 - Chiang Mai, Thailand
Duration: 2014 Dec 92014 Dec 12

Other

Other2014 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA 2014
CountryThailand
CityChiang Mai
Period14/12/914/12/12

Fingerprint

Correlation methods
Face recognition
Adaptive boosting
Image classification
Processing
Experiments

ASJC Scopus subject areas

  • Signal Processing
  • Information Systems

Cite this

Saegusa, C., Yamaoka, M., & Watanabe, K. (2014). Seeing faces in noise: Exploring machine and human face detection processes by the reverse correlation method. In 2014 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA 2014 [7041601] Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/APSIPA.2014.7041601

Seeing faces in noise : Exploring machine and human face detection processes by the reverse correlation method. / Saegusa, Chihiro; Yamaoka, Megumi; Watanabe, Katsumi.

2014 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA 2014. Institute of Electrical and Electronics Engineers Inc., 2014. 7041601.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Saegusa, C, Yamaoka, M & Watanabe, K 2014, Seeing faces in noise: Exploring machine and human face detection processes by the reverse correlation method. in 2014 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA 2014., 7041601, Institute of Electrical and Electronics Engineers Inc., 2014 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA 2014, Chiang Mai, Thailand, 14/12/9. https://doi.org/10.1109/APSIPA.2014.7041601
Saegusa C, Yamaoka M, Watanabe K. Seeing faces in noise: Exploring machine and human face detection processes by the reverse correlation method. In 2014 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA 2014. Institute of Electrical and Electronics Engineers Inc. 2014. 7041601 https://doi.org/10.1109/APSIPA.2014.7041601
Saegusa, Chihiro ; Yamaoka, Megumi ; Watanabe, Katsumi. / Seeing faces in noise : Exploring machine and human face detection processes by the reverse correlation method. 2014 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA 2014. Institute of Electrical and Electronics Engineers Inc., 2014.
@inproceedings{b8041adc9284486ba78f67afd3953f88,
title = "Seeing faces in noise: Exploring machine and human face detection processes by the reverse correlation method",
abstract = "In the present study, we aimed at investigating possible similarities (and discrepancies) between two major machine algorithms of face detection (AdaBoost and EigenFace) and human face detection processes. For this, we presented the 'false classification images' produced by the two face detection algorithms to human observers. Noise fields were fed into the two algorithms and images in which each algorithm falsely detected faces were collected. Those images were averaged and normalized to obtain false classification images. Human observers performed a psychophysical experiment to detect a face with the false classification images against random noise images. The face detection performance increased almost linearly as the number of averaged false detection images increase. Inverted images reduced the detection performance more with the images produced by EigenFace than those by AdaBoost. The present results suggest that both human and machine detection algorithms tended to make similar errors and therefore both AdaBoost and EigenFace are good approximation of human face processing.",
author = "Chihiro Saegusa and Megumi Yamaoka and Katsumi Watanabe",
year = "2014",
month = "2",
day = "12",
doi = "10.1109/APSIPA.2014.7041601",
language = "English",
booktitle = "2014 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA 2014",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
address = "United States",

}

TY - GEN

T1 - Seeing faces in noise

T2 - Exploring machine and human face detection processes by the reverse correlation method

AU - Saegusa, Chihiro

AU - Yamaoka, Megumi

AU - Watanabe, Katsumi

PY - 2014/2/12

Y1 - 2014/2/12

N2 - In the present study, we aimed at investigating possible similarities (and discrepancies) between two major machine algorithms of face detection (AdaBoost and EigenFace) and human face detection processes. For this, we presented the 'false classification images' produced by the two face detection algorithms to human observers. Noise fields were fed into the two algorithms and images in which each algorithm falsely detected faces were collected. Those images were averaged and normalized to obtain false classification images. Human observers performed a psychophysical experiment to detect a face with the false classification images against random noise images. The face detection performance increased almost linearly as the number of averaged false detection images increase. Inverted images reduced the detection performance more with the images produced by EigenFace than those by AdaBoost. The present results suggest that both human and machine detection algorithms tended to make similar errors and therefore both AdaBoost and EigenFace are good approximation of human face processing.

AB - In the present study, we aimed at investigating possible similarities (and discrepancies) between two major machine algorithms of face detection (AdaBoost and EigenFace) and human face detection processes. For this, we presented the 'false classification images' produced by the two face detection algorithms to human observers. Noise fields were fed into the two algorithms and images in which each algorithm falsely detected faces were collected. Those images were averaged and normalized to obtain false classification images. Human observers performed a psychophysical experiment to detect a face with the false classification images against random noise images. The face detection performance increased almost linearly as the number of averaged false detection images increase. Inverted images reduced the detection performance more with the images produced by EigenFace than those by AdaBoost. The present results suggest that both human and machine detection algorithms tended to make similar errors and therefore both AdaBoost and EigenFace are good approximation of human face processing.

UR - http://www.scopus.com/inward/record.url?scp=84983160340&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84983160340&partnerID=8YFLogxK

U2 - 10.1109/APSIPA.2014.7041601

DO - 10.1109/APSIPA.2014.7041601

M3 - Conference contribution

AN - SCOPUS:84983160340

BT - 2014 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference, APSIPA 2014

PB - Institute of Electrical and Electronics Engineers Inc.

ER -