Real-time Activity Recognition by Discerning Qualitative Relationships Between Randomly Chosen Visual Features

Ardhendu Behera, Anthony Cohn, David Hogg

Research output: Contribution to conferencePaper

1 Citation (Scopus)

Abstract

In this paper, we present a novel method to explore semantically meaningful visual information and identify the discriminative spatiotemporal relationships between them for real-time activity recognition. Our approach infers human activities using continuous egocentric (first-person-view) videos of object manipulations in an industrial setup. In order to achieve this goal, we propose a random forest that unifies randomization, discriminative relationships mining and a Markov temporal structure. Discriminative relationships mining helps us to model relations that distinguish different activities, while randomization allows us to handle the large feature space and prevents over-fitting. The Markov temporal structure provides temporally consistent decisions during testing. The proposed random forest uses a discriminative Markov decision tree, where every nonterminal node is a discriminative classifier and the Markov structure is applied at leaf nodes. The proposed approach outperforms the state-of-the-art methods on a new challenging video dataset of assembling a pump system.
Original languageEnglish
Pages1-13
Publication statusPublished - 1 Sep 2014
Event25th British Machine Vision Conference - Nottingham, United Kingdom
Duration: 1 Sep 20145 Sep 2014

Conference

Conference25th British Machine Vision Conference
CountryUnited Kingdom
CityNottingham
Period1/09/145/09/14

Fingerprint

Decision trees
Classifiers
Pumps
Testing

Cite this

Behera, A., Cohn, A., & Hogg, D. (2014). Real-time Activity Recognition by Discerning Qualitative Relationships Between Randomly Chosen Visual Features. 1-13. Paper presented at 25th British Machine Vision Conference, Nottingham, United Kingdom.
Behera, Ardhendu ; Cohn, Anthony ; Hogg, David. / Real-time Activity Recognition by Discerning Qualitative Relationships Between Randomly Chosen Visual Features. Paper presented at 25th British Machine Vision Conference, Nottingham, United Kingdom.
@conference{5b8e2088dda44d649de9d8467ed1f713,
title = "Real-time Activity Recognition by Discerning Qualitative Relationships Between Randomly Chosen Visual Features",
abstract = "In this paper, we present a novel method to explore semantically meaningful visual information and identify the discriminative spatiotemporal relationships between them for real-time activity recognition. Our approach infers human activities using continuous egocentric (first-person-view) videos of object manipulations in an industrial setup. In order to achieve this goal, we propose a random forest that unifies randomization, discriminative relationships mining and a Markov temporal structure. Discriminative relationships mining helps us to model relations that distinguish different activities, while randomization allows us to handle the large feature space and prevents over-fitting. The Markov temporal structure provides temporally consistent decisions during testing. The proposed random forest uses a discriminative Markov decision tree, where every nonterminal node is a discriminative classifier and the Markov structure is applied at leaf nodes. The proposed approach outperforms the state-of-the-art methods on a new challenging video dataset of assembling a pump system.",
author = "Ardhendu Behera and Anthony Cohn and David Hogg",
year = "2014",
month = "9",
day = "1",
language = "English",
pages = "1--13",
note = "25th British Machine Vision Conference ; Conference date: 01-09-2014 Through 05-09-2014",

}

Behera, A, Cohn, A & Hogg, D 2014, 'Real-time Activity Recognition by Discerning Qualitative Relationships Between Randomly Chosen Visual Features' Paper presented at 25th British Machine Vision Conference, Nottingham, United Kingdom, 1/09/14 - 5/09/14, pp. 1-13.

Real-time Activity Recognition by Discerning Qualitative Relationships Between Randomly Chosen Visual Features. / Behera, Ardhendu; Cohn, Anthony; Hogg, David.

2014. 1-13 Paper presented at 25th British Machine Vision Conference, Nottingham, United Kingdom.

Research output: Contribution to conferencePaper

TY - CONF

T1 - Real-time Activity Recognition by Discerning Qualitative Relationships Between Randomly Chosen Visual Features

AU - Behera, Ardhendu

AU - Cohn, Anthony

AU - Hogg, David

PY - 2014/9/1

Y1 - 2014/9/1

N2 - In this paper, we present a novel method to explore semantically meaningful visual information and identify the discriminative spatiotemporal relationships between them for real-time activity recognition. Our approach infers human activities using continuous egocentric (first-person-view) videos of object manipulations in an industrial setup. In order to achieve this goal, we propose a random forest that unifies randomization, discriminative relationships mining and a Markov temporal structure. Discriminative relationships mining helps us to model relations that distinguish different activities, while randomization allows us to handle the large feature space and prevents over-fitting. The Markov temporal structure provides temporally consistent decisions during testing. The proposed random forest uses a discriminative Markov decision tree, where every nonterminal node is a discriminative classifier and the Markov structure is applied at leaf nodes. The proposed approach outperforms the state-of-the-art methods on a new challenging video dataset of assembling a pump system.

AB - In this paper, we present a novel method to explore semantically meaningful visual information and identify the discriminative spatiotemporal relationships between them for real-time activity recognition. Our approach infers human activities using continuous egocentric (first-person-view) videos of object manipulations in an industrial setup. In order to achieve this goal, we propose a random forest that unifies randomization, discriminative relationships mining and a Markov temporal structure. Discriminative relationships mining helps us to model relations that distinguish different activities, while randomization allows us to handle the large feature space and prevents over-fitting. The Markov temporal structure provides temporally consistent decisions during testing. The proposed random forest uses a discriminative Markov decision tree, where every nonterminal node is a discriminative classifier and the Markov structure is applied at leaf nodes. The proposed approach outperforms the state-of-the-art methods on a new challenging video dataset of assembling a pump system.

UR - http://www.bmva.org/bmvc/2014/

UR - http://bmvc2014.cs.nott.ac.uk/

M3 - Paper

SP - 1

EP - 13

ER -

Behera A, Cohn A, Hogg D. Real-time Activity Recognition by Discerning Qualitative Relationships Between Randomly Chosen Visual Features. 2014. Paper presented at 25th British Machine Vision Conference, Nottingham, United Kingdom.