Detection, alignment, and recognition of partially occluded human face based on statistical learning of local features

Jeongin Seo; Hyeyoung Park

doi:10.1145/3055635.3056612

Detection, alignment, and recognition of partially occluded human face based on statistical learning of local features

Jeongin Seo, Hyeyoung Park

Kyungpook National University

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

1 Scopus citations

Abstract

This paper proposes an integrated method to detect, align, and recognize human face focusing on maintaining its stability under partial occlusions. The proposed framework is based on the combination of three probabilistic models: face model, variation model, and background model. These probabilistic models are estimated by analyzing statistical distribution of local features from human face images and natural scene images. By combining three models, it is expected to achieve robustness to various local variations including partial occlusions. Through comparison with well-known face detection algorithm and a recent deep learning architecture, we confirm the robustness of proposed method in detection and recognition of occluded faces.

Original language	English
Title of host publication	Proceedings of 2017 9th International Conference on Machine Learning and Computing, ICMLC 2017
Publisher	Association for Computing Machinery
Pages	289-293
Number of pages	5
ISBN (Electronic)	9781450348171
DOIs	https://doi.org/10.1145/3055635.3056612
State	Published - 24 Feb 2017
Event	9th International Conference on Machine Learning and Computing, ICMLC 2017 - Singapore, Singapore Duration: 24 Feb 2017 → 26 Feb 2017

Publication series

Name	ACM International Conference Proceeding Series
Volume	Part F128357

Conference

Conference	9th International Conference on Machine Learning and Computing, ICMLC 2017
Country/Territory	Singapore
City	Singapore
Period	24/02/17 → 26/02/17

Keywords

Face alignment
Face detection
Face recognition
Local feature
Partial occlusion
Statistical learning

Access to Document

10.1145/3055635.3056612

Cite this

Seo, J., & Park, H. (2017). Detection, alignment, and recognition of partially occluded human face based on statistical learning of local features. In Proceedings of 2017 9th International Conference on Machine Learning and Computing, ICMLC 2017 (pp. 289-293). (ACM International Conference Proceeding Series; Vol. Part F128357). Association for Computing Machinery. https://doi.org/10.1145/3055635.3056612

@inproceedings{500b2af6133b40e79a8bb0bd7e543967,

title = "Detection, alignment, and recognition of partially occluded human face based on statistical learning of local features",

abstract = "This paper proposes an integrated method to detect, align, and recognize human face focusing on maintaining its stability under partial occlusions. The proposed framework is based on the combination of three probabilistic models: face model, variation model, and background model. These probabilistic models are estimated by analyzing statistical distribution of local features from human face images and natural scene images. By combining three models, it is expected to achieve robustness to various local variations including partial occlusions. Through comparison with well-known face detection algorithm and a recent deep learning architecture, we confirm the robustness of proposed method in detection and recognition of occluded faces.",

keywords = "Face alignment, Face detection, Face recognition, Local feature, Partial occlusion, Statistical learning",

author = "Jeongin Seo and Hyeyoung Park",

note = "Publisher Copyright: {\textcopyright} 2017 ACM.; 9th International Conference on Machine Learning and Computing, ICMLC 2017 ; Conference date: 24-02-2017 Through 26-02-2017",

year = "2017",

month = feb,

day = "24",

doi = "10.1145/3055635.3056612",

language = "English",

series = "ACM International Conference Proceeding Series",

publisher = "Association for Computing Machinery",

pages = "289--293",

booktitle = "Proceedings of 2017 9th International Conference on Machine Learning and Computing, ICMLC 2017",

}

Seo, J & Park, H 2017, Detection, alignment, and recognition of partially occluded human face based on statistical learning of local features. in Proceedings of 2017 9th International Conference on Machine Learning and Computing, ICMLC 2017. ACM International Conference Proceeding Series, vol. Part F128357, Association for Computing Machinery, pp. 289-293, 9th International Conference on Machine Learning and Computing, ICMLC 2017, Singapore, Singapore, 24/02/17. https://doi.org/10.1145/3055635.3056612

Detection, alignment, and recognition of partially occluded human face based on statistical learning of local features. / Seo, Jeongin; Park, Hyeyoung.
Proceedings of 2017 9th International Conference on Machine Learning and Computing, ICMLC 2017. Association for Computing Machinery, 2017. p. 289-293 (ACM International Conference Proceeding Series; Vol. Part F128357).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Detection, alignment, and recognition of partially occluded human face based on statistical learning of local features

AU - Seo, Jeongin

AU - Park, Hyeyoung

PY - 2017/2/24

Y1 - 2017/2/24

N2 - This paper proposes an integrated method to detect, align, and recognize human face focusing on maintaining its stability under partial occlusions. The proposed framework is based on the combination of three probabilistic models: face model, variation model, and background model. These probabilistic models are estimated by analyzing statistical distribution of local features from human face images and natural scene images. By combining three models, it is expected to achieve robustness to various local variations including partial occlusions. Through comparison with well-known face detection algorithm and a recent deep learning architecture, we confirm the robustness of proposed method in detection and recognition of occluded faces.

AB - This paper proposes an integrated method to detect, align, and recognize human face focusing on maintaining its stability under partial occlusions. The proposed framework is based on the combination of three probabilistic models: face model, variation model, and background model. These probabilistic models are estimated by analyzing statistical distribution of local features from human face images and natural scene images. By combining three models, it is expected to achieve robustness to various local variations including partial occlusions. Through comparison with well-known face detection algorithm and a recent deep learning architecture, we confirm the robustness of proposed method in detection and recognition of occluded faces.

KW - Face alignment

KW - Face detection

KW - Face recognition

KW - Local feature

KW - Partial occlusion

KW - Statistical learning

UR - http://www.scopus.com/inward/record.url?scp=85025151338&partnerID=8YFLogxK

U2 - 10.1145/3055635.3056612

DO - 10.1145/3055635.3056612

M3 - Conference contribution

AN - SCOPUS:85025151338

T3 - ACM International Conference Proceeding Series

SP - 289

EP - 293

BT - Proceedings of 2017 9th International Conference on Machine Learning and Computing, ICMLC 2017

PB - Association for Computing Machinery

T2 - 9th International Conference on Machine Learning and Computing, ICMLC 2017

Y2 - 24 February 2017 through 26 February 2017

ER -

Detection, alignment, and recognition of partially occluded human face based on statistical learning of local features

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this