Reliable attribute-based object recognition using high predictive value classifiers

Wentao Luan, Yezhou Yang, Cornelia Fermüller, John S. Baras

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

Abstract

We consider the problem of object recognition in 3D using an ensemble of attribute-based classifiers. We propose two new concepts to improve classification in practical situations, and show their implementation in an approach implemented for recognition from point-cloud data. First, the viewing conditions can have a strong influence on classification performance. We study the impact of the distance between the camera and the object and propose an approach to fusing multiple attribute classifiers, which incorporates distance into the decision making. Second, lack of representative training samples often makes it difficult to learn the optimal threshold value for best positive and negative detection rate. We address this issue, by setting in our attribute classifiers instead of just one threshold value, two threshold values to distinguish a positive, a negative and an uncertainty class, and we prove the theoretical correctness of this approach. Empirical studies demonstrate the effectiveness and feasibility of the proposed concepts.

Original languageEnglish (US)
Title of host publicationComputer Vision - 14th European Conference, ECCV 2016, Proceedings
PublisherSpringer Verlag
Pages801-815
Number of pages15
Volume9907 LNCS
ISBN (Print)9783319464862
DOIs
StatePublished - 2016
Externally publishedYes

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume9907 LNCS
ISSN (Print)03029743
ISSN (Electronic)16113349

Fingerprint

Object recognition
Object Recognition
Threshold Value
Classifiers
Classifier
Attribute
Point Cloud
Training Samples
Empirical Study
Correctness
Ensemble
Decision making
Camera
Decision Making
Cameras
Uncertainty
Demonstrate
Concepts

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Luan, W., Yang, Y., Fermüller, C., & Baras, J. S. (2016). Reliable attribute-based object recognition using high predictive value classifiers. In Computer Vision - 14th European Conference, ECCV 2016, Proceedings (Vol. 9907 LNCS, pp. 801-815). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9907 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-319-46487-9_49

Reliable attribute-based object recognition using high predictive value classifiers. / Luan, Wentao; Yang, Yezhou; Fermüller, Cornelia; Baras, John S.

Computer Vision - 14th European Conference, ECCV 2016, Proceedings. Vol. 9907 LNCS Springer Verlag, 2016. p. 801-815 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9907 LNCS).

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

Luan, W, Yang, Y, Fermüller, C & Baras, JS 2016, Reliable attribute-based object recognition using high predictive value classifiers. in Computer Vision - 14th European Conference, ECCV 2016, Proceedings. vol. 9907 LNCS, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9907 LNCS, Springer Verlag, pp. 801-815. https://doi.org/10.1007/978-3-319-46487-9_49
Luan W, Yang Y, Fermüller C, Baras JS. Reliable attribute-based object recognition using high predictive value classifiers. In Computer Vision - 14th European Conference, ECCV 2016, Proceedings. Vol. 9907 LNCS. Springer Verlag. 2016. p. 801-815. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-319-46487-9_49
Luan, Wentao ; Yang, Yezhou ; Fermüller, Cornelia ; Baras, John S. / Reliable attribute-based object recognition using high predictive value classifiers. Computer Vision - 14th European Conference, ECCV 2016, Proceedings. Vol. 9907 LNCS Springer Verlag, 2016. pp. 801-815 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{2d96d92232b54e468a681c1e4e7f6499,
title = "Reliable attribute-based object recognition using high predictive value classifiers",
abstract = "We consider the problem of object recognition in 3D using an ensemble of attribute-based classifiers. We propose two new concepts to improve classification in practical situations, and show their implementation in an approach implemented for recognition from point-cloud data. First, the viewing conditions can have a strong influence on classification performance. We study the impact of the distance between the camera and the object and propose an approach to fusing multiple attribute classifiers, which incorporates distance into the decision making. Second, lack of representative training samples often makes it difficult to learn the optimal threshold value for best positive and negative detection rate. We address this issue, by setting in our attribute classifiers instead of just one threshold value, two threshold values to distinguish a positive, a negative and an uncertainty class, and we prove the theoretical correctness of this approach. Empirical studies demonstrate the effectiveness and feasibility of the proposed concepts.",
author = "Wentao Luan and Yezhou Yang and Cornelia Ferm{\"u}ller and Baras, {John S.}",
year = "2016",
doi = "10.1007/978-3-319-46487-9_49",
language = "English (US)",
isbn = "9783319464862",
volume = "9907 LNCS",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "801--815",
booktitle = "Computer Vision - 14th European Conference, ECCV 2016, Proceedings",
address = "Germany",

}

TY - GEN

T1 - Reliable attribute-based object recognition using high predictive value classifiers

AU - Luan, Wentao

AU - Yang, Yezhou

AU - Fermüller, Cornelia

AU - Baras, John S.

PY - 2016

Y1 - 2016

N2 - We consider the problem of object recognition in 3D using an ensemble of attribute-based classifiers. We propose two new concepts to improve classification in practical situations, and show their implementation in an approach implemented for recognition from point-cloud data. First, the viewing conditions can have a strong influence on classification performance. We study the impact of the distance between the camera and the object and propose an approach to fusing multiple attribute classifiers, which incorporates distance into the decision making. Second, lack of representative training samples often makes it difficult to learn the optimal threshold value for best positive and negative detection rate. We address this issue, by setting in our attribute classifiers instead of just one threshold value, two threshold values to distinguish a positive, a negative and an uncertainty class, and we prove the theoretical correctness of this approach. Empirical studies demonstrate the effectiveness and feasibility of the proposed concepts.

AB - We consider the problem of object recognition in 3D using an ensemble of attribute-based classifiers. We propose two new concepts to improve classification in practical situations, and show their implementation in an approach implemented for recognition from point-cloud data. First, the viewing conditions can have a strong influence on classification performance. We study the impact of the distance between the camera and the object and propose an approach to fusing multiple attribute classifiers, which incorporates distance into the decision making. Second, lack of representative training samples often makes it difficult to learn the optimal threshold value for best positive and negative detection rate. We address this issue, by setting in our attribute classifiers instead of just one threshold value, two threshold values to distinguish a positive, a negative and an uncertainty class, and we prove the theoretical correctness of this approach. Empirical studies demonstrate the effectiveness and feasibility of the proposed concepts.

UR - http://www.scopus.com/inward/record.url?scp=84990041919&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84990041919&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-46487-9_49

DO - 10.1007/978-3-319-46487-9_49

M3 - Conference contribution

AN - SCOPUS:84990041919

SN - 9783319464862

VL - 9907 LNCS

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

SP - 801

EP - 815

BT - Computer Vision - 14th European Conference, ECCV 2016, Proceedings

PB - Springer Verlag

ER -