Subset feature learning for fine-grained category classification

ZongYuan Ge, Christopher McCool, Conrad Sanderson, Peter Corke

Research output: Chapter in Book/Report/Conference proceedingConference PaperResearchpeer-review

18 Citations (Scopus)

Abstract

Fine-grained categorisation has been a challenging problem due to small inter-class variation, large intra-class variation and low number of training images. We propose a learning system which first clusters visually similar classes and then learns deep convolutional neural network features specific to each subset. Experiments on the popular fine-grained Caltech-UCSD bird dataset show that the proposed method outperforms recent fine-grained categorisation methods under the most difficult setting: no bounding boxes are presented at test time. It achieves a mean accuracy of 77.5%, compared to the previous best performance of 73.2%. We also show that progressive transfer learning allows us to first learn domain-generic features (for bird classification) which can then be adapted to specific set of bird classes, yielding improvements in accuracy.

Original languageEnglish
Title of host publication2015 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW 2015)
EditorsKristen Grauman, Erik Learned-Miller, Antonio Torralba, Andrew Zisserman
Place of PublicationPiscataway NJ USA
PublisherIEEE, Institute of Electrical and Electronics Engineers
Pages46-52
Number of pages7
ISBN (Print)9781467367608
DOIs
Publication statusPublished - 19 Oct 2015
Externally publishedYes
EventIEEE Conference on Computer Vision and Pattern Recognition 2015 - Hynes Convention Center, Boston, United States of America
Duration: 7 Jun 201512 Jun 2015
http://www.pamitc.org/cvpr15/

Conference

ConferenceIEEE Conference on Computer Vision and Pattern Recognition 2015
Abbreviated titleCVPR 2015
CountryUnited States of America
CityBoston
Period7/06/1512/06/15
Internet address

Keywords

  • Accuracy
  • Australia
  • Birds
  • Feature extraction
  • Learning systems
  • Neural networks
  • Training

Cite this

Ge, Z., McCool, C., Sanderson, C., & Corke, P. (2015). Subset feature learning for fine-grained category classification. In K. Grauman, E. Learned-Miller, A. Torralba, & A. Zisserman (Eds.), 2015 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW 2015) (pp. 46-52). [7301271] Piscataway NJ USA: IEEE, Institute of Electrical and Electronics Engineers. https://doi.org/10.1109/CVPRW.2015.7301271
Ge, ZongYuan ; McCool, Christopher ; Sanderson, Conrad ; Corke, Peter. / Subset feature learning for fine-grained category classification. 2015 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW 2015). editor / Kristen Grauman ; Erik Learned-Miller ; Antonio Torralba ; Andrew Zisserman. Piscataway NJ USA : IEEE, Institute of Electrical and Electronics Engineers, 2015. pp. 46-52
@inproceedings{3b22019f357c428aa18b8d3cd2189ac8,
title = "Subset feature learning for fine-grained category classification",
abstract = "Fine-grained categorisation has been a challenging problem due to small inter-class variation, large intra-class variation and low number of training images. We propose a learning system which first clusters visually similar classes and then learns deep convolutional neural network features specific to each subset. Experiments on the popular fine-grained Caltech-UCSD bird dataset show that the proposed method outperforms recent fine-grained categorisation methods under the most difficult setting: no bounding boxes are presented at test time. It achieves a mean accuracy of 77.5{\%}, compared to the previous best performance of 73.2{\%}. We also show that progressive transfer learning allows us to first learn domain-generic features (for bird classification) which can then be adapted to specific set of bird classes, yielding improvements in accuracy.",
keywords = "Accuracy, Australia, Birds, Feature extraction, Learning systems, Neural networks, Training",
author = "ZongYuan Ge and Christopher McCool and Conrad Sanderson and Peter Corke",
year = "2015",
month = "10",
day = "19",
doi = "10.1109/CVPRW.2015.7301271",
language = "English",
isbn = "9781467367608",
pages = "46--52",
editor = "Kristen Grauman and Erik Learned-Miller and Antonio Torralba and Andrew Zisserman",
booktitle = "2015 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW 2015)",
publisher = "IEEE, Institute of Electrical and Electronics Engineers",
address = "United States of America",

}

Ge, Z, McCool, C, Sanderson, C & Corke, P 2015, Subset feature learning for fine-grained category classification. in K Grauman, E Learned-Miller, A Torralba & A Zisserman (eds), 2015 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW 2015)., 7301271, IEEE, Institute of Electrical and Electronics Engineers, Piscataway NJ USA, pp. 46-52, IEEE Conference on Computer Vision and Pattern Recognition 2015, Boston, United States of America, 7/06/15. https://doi.org/10.1109/CVPRW.2015.7301271

Subset feature learning for fine-grained category classification. / Ge, ZongYuan; McCool, Christopher; Sanderson, Conrad; Corke, Peter.

2015 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW 2015). ed. / Kristen Grauman; Erik Learned-Miller; Antonio Torralba; Andrew Zisserman. Piscataway NJ USA : IEEE, Institute of Electrical and Electronics Engineers, 2015. p. 46-52 7301271.

Research output: Chapter in Book/Report/Conference proceedingConference PaperResearchpeer-review

TY - GEN

T1 - Subset feature learning for fine-grained category classification

AU - Ge, ZongYuan

AU - McCool, Christopher

AU - Sanderson, Conrad

AU - Corke, Peter

PY - 2015/10/19

Y1 - 2015/10/19

N2 - Fine-grained categorisation has been a challenging problem due to small inter-class variation, large intra-class variation and low number of training images. We propose a learning system which first clusters visually similar classes and then learns deep convolutional neural network features specific to each subset. Experiments on the popular fine-grained Caltech-UCSD bird dataset show that the proposed method outperforms recent fine-grained categorisation methods under the most difficult setting: no bounding boxes are presented at test time. It achieves a mean accuracy of 77.5%, compared to the previous best performance of 73.2%. We also show that progressive transfer learning allows us to first learn domain-generic features (for bird classification) which can then be adapted to specific set of bird classes, yielding improvements in accuracy.

AB - Fine-grained categorisation has been a challenging problem due to small inter-class variation, large intra-class variation and low number of training images. We propose a learning system which first clusters visually similar classes and then learns deep convolutional neural network features specific to each subset. Experiments on the popular fine-grained Caltech-UCSD bird dataset show that the proposed method outperforms recent fine-grained categorisation methods under the most difficult setting: no bounding boxes are presented at test time. It achieves a mean accuracy of 77.5%, compared to the previous best performance of 73.2%. We also show that progressive transfer learning allows us to first learn domain-generic features (for bird classification) which can then be adapted to specific set of bird classes, yielding improvements in accuracy.

KW - Accuracy

KW - Australia

KW - Birds

KW - Feature extraction

KW - Learning systems

KW - Neural networks

KW - Training

UR - http://www.scopus.com/inward/record.url?scp=84951950853&partnerID=8YFLogxK

U2 - 10.1109/CVPRW.2015.7301271

DO - 10.1109/CVPRW.2015.7301271

M3 - Conference Paper

SN - 9781467367608

SP - 46

EP - 52

BT - 2015 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW 2015)

A2 - Grauman, Kristen

A2 - Learned-Miller, Erik

A2 - Torralba, Antonio

A2 - Zisserman, Andrew

PB - IEEE, Institute of Electrical and Electronics Engineers

CY - Piscataway NJ USA

ER -

Ge Z, McCool C, Sanderson C, Corke P. Subset feature learning for fine-grained category classification. In Grauman K, Learned-Miller E, Torralba A, Zisserman A, editors, 2015 IEEE Conference on Computer Vision and Pattern Recognition Workshops (CVPRW 2015). Piscataway NJ USA: IEEE, Institute of Electrical and Electronics Engineers. 2015. p. 46-52. 7301271 https://doi.org/10.1109/CVPRW.2015.7301271