Temporal self-similarity for appearance-based action recognition in multi-view setups

Marco Körner; Joachim Denzler

doi:10.1007/978-3-642-40261-6_19

Temporal self-similarity for appearance-based action recognition in multi-view setups

Marco Körner, Joachim Denzler

Friedrich Schiller University Jena

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

9 Scopus citations

Abstract

We present a general data-driven method for multi-view action recognition relying on the appearance of dynamic systems captured from different viewpoints. Thus, we do not depend on 3d reconstruction, foreground segmentation, or accurate detections. We extend further earlier approaches based on Temporal Self-Similarity Maps by new low-level image features and similarity measures. Gaussian Process classification in combination with Histogram Intersection Kernels serve as powerful tools in our approach. Experiments performed on our new combined multi-view dataset as well as on the widely used IXMAS dataset show promising and competing results.

Original language	English
Title of host publication	Computer Analysis of Images and Patterns - 15th International Conference, CAIP 2013, Proceedings
Pages	163-171
Number of pages	9
Edition	PART 1
DOIs	https://doi.org/10.1007/978-3-642-40261-6_19
State	Published - 2013
Externally published	Yes
Event	15th International Conference on Computer Analysis of Images and Patterns, CAIP 2013 - York, United Kingdom Duration: 27 Aug 2013 → 29 Aug 2013

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Number	PART 1
Volume	8047 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	15th International Conference on Computer Analysis of Images and Patterns, CAIP 2013
Country/Territory	United Kingdom
City	York
Period	27/08/13 → 29/08/13

Keywords

Action Recognition
Gaussian Processes
Histogram-Intersection Kernel
Multi-View
Temporal Self-Similarity

Access to Document

10.1007/978-3-642-40261-6_19

Cite this

Körner, M., & Denzler, J. (2013). Temporal self-similarity for appearance-based action recognition in multi-view setups. In Computer Analysis of Images and Patterns - 15th International Conference, CAIP 2013, Proceedings (PART 1 ed., pp. 163-171). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8047 LNCS, No. PART 1). https://doi.org/10.1007/978-3-642-40261-6_19

Körner, Marco ; Denzler, Joachim. / Temporal self-similarity for appearance-based action recognition in multi-view setups. Computer Analysis of Images and Patterns - 15th International Conference, CAIP 2013, Proceedings. PART 1. ed. 2013. pp. 163-171 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 1).

@inproceedings{572957c272564c9aa20b4b731f858335,

title = "Temporal self-similarity for appearance-based action recognition in multi-view setups",

abstract = "We present a general data-driven method for multi-view action recognition relying on the appearance of dynamic systems captured from different viewpoints. Thus, we do not depend on 3d reconstruction, foreground segmentation, or accurate detections. We extend further earlier approaches based on Temporal Self-Similarity Maps by new low-level image features and similarity measures. Gaussian Process classification in combination with Histogram Intersection Kernels serve as powerful tools in our approach. Experiments performed on our new combined multi-view dataset as well as on the widely used IXMAS dataset show promising and competing results.",

keywords = "Action Recognition, Gaussian Processes, Histogram-Intersection Kernel, Multi-View, Temporal Self-Similarity",

author = "Marco K{\"o}rner and Joachim Denzler",

year = "2013",

doi = "10.1007/978-3-642-40261-6_19",

language = "English",

isbn = "9783642402609",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

number = "PART 1",

pages = "163--171",

booktitle = "Computer Analysis of Images and Patterns - 15th International Conference, CAIP 2013, Proceedings",

edition = "PART 1",

note = "15th International Conference on Computer Analysis of Images and Patterns, CAIP 2013 ; Conference date: 27-08-2013 Through 29-08-2013",

}

Körner, M & Denzler, J 2013, Temporal self-similarity for appearance-based action recognition in multi-view setups. in Computer Analysis of Images and Patterns - 15th International Conference, CAIP 2013, Proceedings. PART 1 edn, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 1, vol. 8047 LNCS, pp. 163-171, 15th International Conference on Computer Analysis of Images and Patterns, CAIP 2013, York, United Kingdom, 27/08/13. https://doi.org/10.1007/978-3-642-40261-6_19

Temporal self-similarity for appearance-based action recognition in multi-view setups. / Körner, Marco; Denzler, Joachim.
Computer Analysis of Images and Patterns - 15th International Conference, CAIP 2013, Proceedings. PART 1. ed. 2013. p. 163-171 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 8047 LNCS, No. PART 1).

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

TY - GEN

T1 - Temporal self-similarity for appearance-based action recognition in multi-view setups

AU - Körner, Marco

AU - Denzler, Joachim

PY - 2013

Y1 - 2013

N2 - We present a general data-driven method for multi-view action recognition relying on the appearance of dynamic systems captured from different viewpoints. Thus, we do not depend on 3d reconstruction, foreground segmentation, or accurate detections. We extend further earlier approaches based on Temporal Self-Similarity Maps by new low-level image features and similarity measures. Gaussian Process classification in combination with Histogram Intersection Kernels serve as powerful tools in our approach. Experiments performed on our new combined multi-view dataset as well as on the widely used IXMAS dataset show promising and competing results.

AB - We present a general data-driven method for multi-view action recognition relying on the appearance of dynamic systems captured from different viewpoints. Thus, we do not depend on 3d reconstruction, foreground segmentation, or accurate detections. We extend further earlier approaches based on Temporal Self-Similarity Maps by new low-level image features and similarity measures. Gaussian Process classification in combination with Histogram Intersection Kernels serve as powerful tools in our approach. Experiments performed on our new combined multi-view dataset as well as on the widely used IXMAS dataset show promising and competing results.

KW - Action Recognition

KW - Gaussian Processes

KW - Histogram-Intersection Kernel

KW - Multi-View

KW - Temporal Self-Similarity

UR - http://www.scopus.com/inward/record.url?scp=84884471976&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-40261-6_19

DO - 10.1007/978-3-642-40261-6_19

M3 - Conference contribution

AN - SCOPUS:84884471976

SN - 9783642402609

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 163

EP - 171

BT - Computer Analysis of Images and Patterns - 15th International Conference, CAIP 2013, Proceedings

T2 - 15th International Conference on Computer Analysis of Images and Patterns, CAIP 2013

Y2 - 27 August 2013 through 29 August 2013

ER -

Körner M, Denzler J. Temporal self-similarity for appearance-based action recognition in multi-view setups. In Computer Analysis of Images and Patterns - 15th International Conference, CAIP 2013, Proceedings. PART 1 ed. 2013. p. 163-171. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 1). doi: 10.1007/978-3-642-40261-6_19

Temporal self-similarity for appearance-based action recognition in multi-view setups

Abstract

Publication series

Conference

Keywords

Access to Document

Other files and links

Fingerprint

Cite this