Best practices to train deep models on imbalanced datasets—a case study on animal detection in aerial imagery

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

Abstract

We introduce recommendations to train a Convolutional Neural Network for grid-based detection on a dataset that has a substantial class imbalance. These include curriculum learning, hard negative mining, a special border class, and more. We evaluate the recommendations on the problem of animal detection in aerial images, where we obtain an increase in precision from 9% to 40% at high recalls, compared to state-of-the-art. Data related to this paper are available at: http://doi.org/10.5281/zenodo.609023.

Original languageEnglish
Title of host publicationMachine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2018, Proceedings
EditorsUlf Brefeld, Alice Marascu, Fabio Pinelli, Edward Curry, Brian MacNamee, Neil Hurley, Elizabeth Daly, Michele Berlingerio
PublisherSpringer Verlag
Pages630-634
ISBN (Print)9783030109967
DOIs
Publication statusPublished - 1 Jan 2019
EventEuropean Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML-PKDD 2018 - Dublin, Ireland
Duration: 10 Sep 201814 Sep 2018

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11053 LNAI
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

ConferenceEuropean Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML-PKDD 2018
CountryIreland
CityDublin
Period10/09/1814/09/18

Fingerprint

Best Practice
Curricula
Recommendations
Animals
Antennas
Neural networks
Aerial Image
Mining
Neural Networks
Grid
Evaluate
Model
Class
Imagery
Curriculum
Learning

Keywords

  • Class imbalance
  • Deep learning
  • Unmanned Aerial Vehicles

Cite this

Kellenberger, B., Marcos, D., & Tuia, D. (2019). Best practices to train deep models on imbalanced datasets—a case study on animal detection in aerial imagery. In U. Brefeld, A. Marascu, F. Pinelli, E. Curry, B. MacNamee, N. Hurley, E. Daly, ... M. Berlingerio (Eds.), Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2018, Proceedings (pp. 630-634). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11053 LNAI). Springer Verlag. https://doi.org/10.1007/978-3-030-10997-4_40
Kellenberger, Benjamin ; Marcos, Diego ; Tuia, Devis. / Best practices to train deep models on imbalanced datasets—a case study on animal detection in aerial imagery. Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2018, Proceedings. editor / Ulf Brefeld ; Alice Marascu ; Fabio Pinelli ; Edward Curry ; Brian MacNamee ; Neil Hurley ; Elizabeth Daly ; Michele Berlingerio. Springer Verlag, 2019. pp. 630-634 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{2170f58434944a40b7698ce0c42f708b,
title = "Best practices to train deep models on imbalanced datasets—a case study on animal detection in aerial imagery",
abstract = "We introduce recommendations to train a Convolutional Neural Network for grid-based detection on a dataset that has a substantial class imbalance. These include curriculum learning, hard negative mining, a special border class, and more. We evaluate the recommendations on the problem of animal detection in aerial images, where we obtain an increase in precision from 9{\%} to 40{\%} at high recalls, compared to state-of-the-art. Data related to this paper are available at: http://doi.org/10.5281/zenodo.609023.",
keywords = "Class imbalance, Deep learning, Unmanned Aerial Vehicles",
author = "Benjamin Kellenberger and Diego Marcos and Devis Tuia",
year = "2019",
month = "1",
day = "1",
doi = "10.1007/978-3-030-10997-4_40",
language = "English",
isbn = "9783030109967",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "630--634",
editor = "Ulf Brefeld and Alice Marascu and Fabio Pinelli and Edward Curry and Brian MacNamee and Neil Hurley and Elizabeth Daly and Michele Berlingerio",
booktitle = "Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2018, Proceedings",

}

Kellenberger, B, Marcos, D & Tuia, D 2019, Best practices to train deep models on imbalanced datasets—a case study on animal detection in aerial imagery. in U Brefeld, A Marascu, F Pinelli, E Curry, B MacNamee, N Hurley, E Daly & M Berlingerio (eds), Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2018, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11053 LNAI, Springer Verlag, pp. 630-634, European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases, ECML-PKDD 2018, Dublin, Ireland, 10/09/18. https://doi.org/10.1007/978-3-030-10997-4_40

Best practices to train deep models on imbalanced datasets—a case study on animal detection in aerial imagery. / Kellenberger, Benjamin; Marcos, Diego; Tuia, Devis.

Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2018, Proceedings. ed. / Ulf Brefeld; Alice Marascu; Fabio Pinelli; Edward Curry; Brian MacNamee; Neil Hurley; Elizabeth Daly; Michele Berlingerio. Springer Verlag, 2019. p. 630-634 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11053 LNAI).

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

TY - GEN

T1 - Best practices to train deep models on imbalanced datasets—a case study on animal detection in aerial imagery

AU - Kellenberger, Benjamin

AU - Marcos, Diego

AU - Tuia, Devis

PY - 2019/1/1

Y1 - 2019/1/1

N2 - We introduce recommendations to train a Convolutional Neural Network for grid-based detection on a dataset that has a substantial class imbalance. These include curriculum learning, hard negative mining, a special border class, and more. We evaluate the recommendations on the problem of animal detection in aerial images, where we obtain an increase in precision from 9% to 40% at high recalls, compared to state-of-the-art. Data related to this paper are available at: http://doi.org/10.5281/zenodo.609023.

AB - We introduce recommendations to train a Convolutional Neural Network for grid-based detection on a dataset that has a substantial class imbalance. These include curriculum learning, hard negative mining, a special border class, and more. We evaluate the recommendations on the problem of animal detection in aerial images, where we obtain an increase in precision from 9% to 40% at high recalls, compared to state-of-the-art. Data related to this paper are available at: http://doi.org/10.5281/zenodo.609023.

KW - Class imbalance

KW - Deep learning

KW - Unmanned Aerial Vehicles

U2 - 10.1007/978-3-030-10997-4_40

DO - 10.1007/978-3-030-10997-4_40

M3 - Conference contribution

SN - 9783030109967

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

SP - 630

EP - 634

BT - Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2018, Proceedings

A2 - Brefeld, Ulf

A2 - Marascu, Alice

A2 - Pinelli, Fabio

A2 - Curry, Edward

A2 - MacNamee, Brian

A2 - Hurley, Neil

A2 - Daly, Elizabeth

A2 - Berlingerio, Michele

PB - Springer Verlag

ER -

Kellenberger B, Marcos D, Tuia D. Best practices to train deep models on imbalanced datasets—a case study on animal detection in aerial imagery. In Brefeld U, Marascu A, Pinelli F, Curry E, MacNamee B, Hurley N, Daly E, Berlingerio M, editors, Machine Learning and Knowledge Discovery in Databases - European Conference, ECML PKDD 2018, Proceedings. Springer Verlag. 2019. p. 630-634. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-030-10997-4_40