Human action recognition by random features and hand-crafted features: a comparative study

Haocheng Shen; Jianguo Zhang; Hui Zhang

doi:10.1007/978-3-319-16181-5_2

Human action recognition by random features and hand-crafted features: a comparative study

Haocheng Shen, Jianguo Zhang, Hui Zhang

Computing

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

4 Citations (Scopus)

Abstract

One popular approach for human action recognition is to extract features from videos as representations, subsequently followed by a classification procedure of the representations. In this paper, we investigate and compare hand-crafted and random feature representation for human action recognition on YouTube dataset. The former is built on 3D HoG/HoF and SIFT descriptors while the latter bases on random projection. Three encoding methods: Bag of Feature(BoF), Sparse Coding(SC) and VLAD are adopted. Spatial temporal pyramid and a twolayer SVM classifier are employed for classification. Our experiments demonstrate that: 1) Sparse Coding is confirmed to outperform Bag of Feature; 2) Using a model of hybrid features incorporating framestatic can significantly improve the overall recognition accuracy; 3) The frame-static features works surprisingly better than motion features only; 4) Compared with the success of hand-crafted feature representation, the random feature representation does not perform well in this dataset.

Original language	English
Title of host publication	Computer Vision - ECCV 2014 Workshops
Subtitle of host publication	Zurich, Switzerland, September 6-7 and 12, 2014, Proceedings, Part II
Editors	Lourdes Agapito , Michael M. Bronstein, Carsten Rother
Publisher	Springer International Publishing
Pages	14-28
Number of pages	15
ISBN (Electronic)	9783319161815
ISBN (Print)	9783319161808
DOIs	https://doi.org/10.1007/978-3-319-16181-5_2
Publication status	Published - 2015
Event	6th International Workshop on Video Event Categorization, Tagging and Retrieval towards Big Data - Zurich, Switzerland Duration: 6 Sept 2014 → 6 Sept 2014 http://eccv2014.org/program_workshops/

Publication series

Name	Lecture notes in computer science
Volume	8926
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Workshop

Workshop	6th International Workshop on Video Event Categorization, Tagging and Retrieval towards Big Data
Abbreviated title	VECTaR 2014
Country/Territory	Switzerland
City	Zurich
Period	6/09/14 → 6/09/14
Other	Part of 13th European Conference on Computer Vision, ECCV 2014
Internet address	http://eccv2014.org/program_workshops/

Keywords

Action recognition
Hand-crafted feature
Random representation

ASJC Scopus subject areas

General Computer Science
Theoretical Computer Science

Access to Document

10.1007/978-3-319-16181-5_2

Cite this

Shen, H., Zhang, J., & Zhang, H. (2015). Human action recognition by random features and hand-crafted features: a comparative study. In L. Agapito , M. M. Bronstein, & C. Rother (Eds.), Computer Vision - ECCV 2014 Workshops: Zurich, Switzerland, September 6-7 and 12, 2014, Proceedings, Part II (pp. 14-28). (Lecture notes in computer science ; Vol. 8926). Springer International Publishing. https://doi.org/10.1007/978-3-319-16181-5_2

Shen, Haocheng ; Zhang, Jianguo ; Zhang, Hui. / Human action recognition by random features and hand-crafted features : a comparative study. Computer Vision - ECCV 2014 Workshops: Zurich, Switzerland, September 6-7 and 12, 2014, Proceedings, Part II. editor / Lourdes Agapito ; Michael M. Bronstein ; Carsten Rother. Springer International Publishing, 2015. pp. 14-28 (Lecture notes in computer science ).

@inproceedings{7c9ebf86b7674404991e5f8b2d3809d7,

title = "Human action recognition by random features and hand-crafted features: a comparative study",

abstract = "One popular approach for human action recognition is to extract features from videos as representations, subsequently followed by a classification procedure of the representations. In this paper, we investigate and compare hand-crafted and random feature representation for human action recognition on YouTube dataset. The former is built on 3D HoG/HoF and SIFT descriptors while the latter bases on random projection. Three encoding methods: Bag of Feature(BoF), Sparse Coding(SC) and VLAD are adopted. Spatial temporal pyramid and a twolayer SVM classifier are employed for classification. Our experiments demonstrate that: 1) Sparse Coding is confirmed to outperform Bag of Feature; 2) Using a model of hybrid features incorporating framestatic can significantly improve the overall recognition accuracy; 3) The frame-static features works surprisingly better than motion features only; 4) Compared with the success of hand-crafted feature representation, the random feature representation does not perform well in this dataset.",

keywords = "Action recognition, Hand-crafted feature, Random representation",

author = "Haocheng Shen and Jianguo Zhang and Hui Zhang",

year = "2015",

doi = "10.1007/978-3-319-16181-5_2",

language = "English",

isbn = "9783319161808",

series = "Lecture notes in computer science ",

publisher = "Springer International Publishing",

pages = "14--28",

editor = "{Agapito }, {Lourdes } and Bronstein, {Michael M. } and Rother, {Carsten }",

booktitle = "Computer Vision - ECCV 2014 Workshops",

address = "Switzerland",

note = "6th International Workshop on Video Event Categorization, Tagging and Retrieval towards Big Data, VECTaR 2014 ; Conference date: 06-09-2014 Through 06-09-2014",

url = "http://eccv2014.org/program_workshops/",

}

Shen, H, Zhang, J & Zhang, H 2015, Human action recognition by random features and hand-crafted features: a comparative study. in L Agapito , MM Bronstein & C Rother (eds), Computer Vision - ECCV 2014 Workshops: Zurich, Switzerland, September 6-7 and 12, 2014, Proceedings, Part II. Lecture notes in computer science , vol. 8926, Springer International Publishing, pp. 14-28, 6th International Workshop on Video Event Categorization, Tagging and Retrieval towards Big Data, Zurich, Switzerland, 6/09/14. https://doi.org/10.1007/978-3-319-16181-5_2

Human action recognition by random features and hand-crafted features: a comparative study. / Shen, Haocheng; Zhang, Jianguo; Zhang, Hui.
Computer Vision - ECCV 2014 Workshops: Zurich, Switzerland, September 6-7 and 12, 2014, Proceedings, Part II. ed. / Lourdes Agapito ; Michael M. Bronstein; Carsten Rother. Springer International Publishing, 2015. p. 14-28 (Lecture notes in computer science ; Vol. 8926).

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

TY - GEN

T1 - Human action recognition by random features and hand-crafted features

T2 - 6th International Workshop on Video Event Categorization, Tagging and Retrieval towards Big Data

AU - Shen, Haocheng

AU - Zhang, Jianguo

AU - Zhang, Hui

PY - 2015

Y1 - 2015

N2 - One popular approach for human action recognition is to extract features from videos as representations, subsequently followed by a classification procedure of the representations. In this paper, we investigate and compare hand-crafted and random feature representation for human action recognition on YouTube dataset. The former is built on 3D HoG/HoF and SIFT descriptors while the latter bases on random projection. Three encoding methods: Bag of Feature(BoF), Sparse Coding(SC) and VLAD are adopted. Spatial temporal pyramid and a twolayer SVM classifier are employed for classification. Our experiments demonstrate that: 1) Sparse Coding is confirmed to outperform Bag of Feature; 2) Using a model of hybrid features incorporating framestatic can significantly improve the overall recognition accuracy; 3) The frame-static features works surprisingly better than motion features only; 4) Compared with the success of hand-crafted feature representation, the random feature representation does not perform well in this dataset.

AB - One popular approach for human action recognition is to extract features from videos as representations, subsequently followed by a classification procedure of the representations. In this paper, we investigate and compare hand-crafted and random feature representation for human action recognition on YouTube dataset. The former is built on 3D HoG/HoF and SIFT descriptors while the latter bases on random projection. Three encoding methods: Bag of Feature(BoF), Sparse Coding(SC) and VLAD are adopted. Spatial temporal pyramid and a twolayer SVM classifier are employed for classification. Our experiments demonstrate that: 1) Sparse Coding is confirmed to outperform Bag of Feature; 2) Using a model of hybrid features incorporating framestatic can significantly improve the overall recognition accuracy; 3) The frame-static features works surprisingly better than motion features only; 4) Compared with the success of hand-crafted feature representation, the random feature representation does not perform well in this dataset.

KW - Action recognition

KW - Hand-crafted feature

KW - Random representation

UR - http://www.scopus.com/inward/record.url?scp=84928787177&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-16181-5_2

DO - 10.1007/978-3-319-16181-5_2

M3 - Conference contribution

AN - SCOPUS:84928787177

SN - 9783319161808

T3 - Lecture notes in computer science

SP - 14

EP - 28

BT - Computer Vision - ECCV 2014 Workshops

A2 - Agapito , Lourdes

A2 - Bronstein, Michael M.

A2 - Rother, Carsten

PB - Springer International Publishing

Y2 - 6 September 2014 through 6 September 2014

ER -

Shen H, Zhang J, Zhang H. Human action recognition by random features and hand-crafted features: a comparative study. In Agapito L, Bronstein MM, Rother C, editors, Computer Vision - ECCV 2014 Workshops: Zurich, Switzerland, September 6-7 and 12, 2014, Proceedings, Part II. Springer International Publishing. 2015. p. 14-28. (Lecture notes in computer science ). doi: 10.1007/978-3-319-16181-5_2

Human action recognition by random features and hand-crafted features: a comparative study

Abstract

Publication series

Workshop

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this