Identifying animal species in camera trap images using deep learning and citizen science

Marco Willi, Ross T. Pitman, Anabelle W. Cardoso, Christina Locke, Alexandra Swanson, Amy Boyer, Marten Veldthuis, Lucy F Fortson

Research output: Contribution to journalArticle

3 Citations (Scopus)

Abstract

Ecologists often study wildlife populations by deploying camera traps. Large datasets are generated using this approach which can be difficult for research teams to manually evaluate. Researchers increasingly enlist volunteers from the general public as citizen scientists to help classify images. The growing number of camera trap studies, however, makes it ever more challenging to find enough volunteers to process all projects in a timely manner. Advances in machine learning, especially deep learning, allow for accurate automatic image classification. By training models using existing datasets of images classified by citizen scientists and subsequent application of such models on new studies, human effort may be reduced substantially. The goals of this study were to (a) assess the accuracy of deep learning in classifying camera trap data, (b) investigate how to process datasets with only a few classified images that are generally difficult to model, and (c) apply a trained model on a live online citizen science project. Convolutional neural networks (CNNs) were used to differentiate among images of different animal species, images of humans or vehicles, and empty images (no animals, vehicles, or humans). We used four different camera trap datasets featuring a wide variety of species, different habitats, and a varying number of images. All datasets were labelled by citizen scientists on Zooniverse. Accuracies for identifying empty images across projects ranged between 91.2% and 98.0%, whereas accuracies for identifying specific species were between 88.7% and 92.7%. Transferring information from CNNs trained on large datasets (“transfer-learning”) was increasingly beneficial as the size of the training dataset decreased and raised accuracy by up to 10.3%. Removing low-confidence predictions increased model accuracies to the level of citizen scientists. By combining a trained model with classifications from citizen scientists, human effort was reduced by 43% while maintaining overall accuracy for a live experiment running on Zooniverse. Ecology researchers can significantly reduce image classification time and manual effort by combining citizen scientists and CNNs, enabling faster processing of data from large camera trap studies.

Original languageEnglish (US)
Pages (from-to)80-91
Number of pages12
JournalMethods in Ecology and Evolution
Volume10
Issue number1
DOIs
StatePublished - Jan 1 2019

Fingerprint

cameras
learning
traps
neural networks
animals
volunteers
image classification
researchers
artificial intelligence
ecologists
animal species
citizen
science
wildlife
taxonomy
ecology
prediction
habitats
animal
habitat

Keywords

  • animal identification
  • camera trap
  • citizen science
  • convolutional neural networks
  • deep learning
  • machine learning

Cite this

Identifying animal species in camera trap images using deep learning and citizen science. / Willi, Marco; Pitman, Ross T.; Cardoso, Anabelle W.; Locke, Christina; Swanson, Alexandra; Boyer, Amy; Veldthuis, Marten; Fortson, Lucy F.

In: Methods in Ecology and Evolution, Vol. 10, No. 1, 01.01.2019, p. 80-91.

Research output: Contribution to journalArticle

Willi, M, Pitman, RT, Cardoso, AW, Locke, C, Swanson, A, Boyer, A, Veldthuis, M & Fortson, LF 2019, 'Identifying animal species in camera trap images using deep learning and citizen science' Methods in Ecology and Evolution, vol. 10, no. 1, pp. 80-91. https://doi.org/10.1111/2041-210X.13099
Willi, Marco ; Pitman, Ross T. ; Cardoso, Anabelle W. ; Locke, Christina ; Swanson, Alexandra ; Boyer, Amy ; Veldthuis, Marten ; Fortson, Lucy F. / Identifying animal species in camera trap images using deep learning and citizen science. In: Methods in Ecology and Evolution. 2019 ; Vol. 10, No. 1. pp. 80-91.
@article{f1d8554edc19434995b7e69c5b2daff2,
title = "Identifying animal species in camera trap images using deep learning and citizen science",
abstract = "Ecologists often study wildlife populations by deploying camera traps. Large datasets are generated using this approach which can be difficult for research teams to manually evaluate. Researchers increasingly enlist volunteers from the general public as citizen scientists to help classify images. The growing number of camera trap studies, however, makes it ever more challenging to find enough volunteers to process all projects in a timely manner. Advances in machine learning, especially deep learning, allow for accurate automatic image classification. By training models using existing datasets of images classified by citizen scientists and subsequent application of such models on new studies, human effort may be reduced substantially. The goals of this study were to (a) assess the accuracy of deep learning in classifying camera trap data, (b) investigate how to process datasets with only a few classified images that are generally difficult to model, and (c) apply a trained model on a live online citizen science project. Convolutional neural networks (CNNs) were used to differentiate among images of different animal species, images of humans or vehicles, and empty images (no animals, vehicles, or humans). We used four different camera trap datasets featuring a wide variety of species, different habitats, and a varying number of images. All datasets were labelled by citizen scientists on Zooniverse. Accuracies for identifying empty images across projects ranged between 91.2{\%} and 98.0{\%}, whereas accuracies for identifying specific species were between 88.7{\%} and 92.7{\%}. Transferring information from CNNs trained on large datasets (“transfer-learning”) was increasingly beneficial as the size of the training dataset decreased and raised accuracy by up to 10.3{\%}. Removing low-confidence predictions increased model accuracies to the level of citizen scientists. By combining a trained model with classifications from citizen scientists, human effort was reduced by 43{\%} while maintaining overall accuracy for a live experiment running on Zooniverse. Ecology researchers can significantly reduce image classification time and manual effort by combining citizen scientists and CNNs, enabling faster processing of data from large camera trap studies.",
keywords = "animal identification, camera trap, citizen science, convolutional neural networks, deep learning, machine learning",
author = "Marco Willi and Pitman, {Ross T.} and Cardoso, {Anabelle W.} and Christina Locke and Alexandra Swanson and Amy Boyer and Marten Veldthuis and Fortson, {Lucy F}",
year = "2019",
month = "1",
day = "1",
doi = "10.1111/2041-210X.13099",
language = "English (US)",
volume = "10",
pages = "80--91",
journal = "Methods in Ecology and Evolution",
issn = "2041-210X",
publisher = "John Wiley and Sons Inc.",
number = "1",

}

TY - JOUR

T1 - Identifying animal species in camera trap images using deep learning and citizen science

AU - Willi, Marco

AU - Pitman, Ross T.

AU - Cardoso, Anabelle W.

AU - Locke, Christina

AU - Swanson, Alexandra

AU - Boyer, Amy

AU - Veldthuis, Marten

AU - Fortson, Lucy F

PY - 2019/1/1

Y1 - 2019/1/1

N2 - Ecologists often study wildlife populations by deploying camera traps. Large datasets are generated using this approach which can be difficult for research teams to manually evaluate. Researchers increasingly enlist volunteers from the general public as citizen scientists to help classify images. The growing number of camera trap studies, however, makes it ever more challenging to find enough volunteers to process all projects in a timely manner. Advances in machine learning, especially deep learning, allow for accurate automatic image classification. By training models using existing datasets of images classified by citizen scientists and subsequent application of such models on new studies, human effort may be reduced substantially. The goals of this study were to (a) assess the accuracy of deep learning in classifying camera trap data, (b) investigate how to process datasets with only a few classified images that are generally difficult to model, and (c) apply a trained model on a live online citizen science project. Convolutional neural networks (CNNs) were used to differentiate among images of different animal species, images of humans or vehicles, and empty images (no animals, vehicles, or humans). We used four different camera trap datasets featuring a wide variety of species, different habitats, and a varying number of images. All datasets were labelled by citizen scientists on Zooniverse. Accuracies for identifying empty images across projects ranged between 91.2% and 98.0%, whereas accuracies for identifying specific species were between 88.7% and 92.7%. Transferring information from CNNs trained on large datasets (“transfer-learning”) was increasingly beneficial as the size of the training dataset decreased and raised accuracy by up to 10.3%. Removing low-confidence predictions increased model accuracies to the level of citizen scientists. By combining a trained model with classifications from citizen scientists, human effort was reduced by 43% while maintaining overall accuracy for a live experiment running on Zooniverse. Ecology researchers can significantly reduce image classification time and manual effort by combining citizen scientists and CNNs, enabling faster processing of data from large camera trap studies.

AB - Ecologists often study wildlife populations by deploying camera traps. Large datasets are generated using this approach which can be difficult for research teams to manually evaluate. Researchers increasingly enlist volunteers from the general public as citizen scientists to help classify images. The growing number of camera trap studies, however, makes it ever more challenging to find enough volunteers to process all projects in a timely manner. Advances in machine learning, especially deep learning, allow for accurate automatic image classification. By training models using existing datasets of images classified by citizen scientists and subsequent application of such models on new studies, human effort may be reduced substantially. The goals of this study were to (a) assess the accuracy of deep learning in classifying camera trap data, (b) investigate how to process datasets with only a few classified images that are generally difficult to model, and (c) apply a trained model on a live online citizen science project. Convolutional neural networks (CNNs) were used to differentiate among images of different animal species, images of humans or vehicles, and empty images (no animals, vehicles, or humans). We used four different camera trap datasets featuring a wide variety of species, different habitats, and a varying number of images. All datasets were labelled by citizen scientists on Zooniverse. Accuracies for identifying empty images across projects ranged between 91.2% and 98.0%, whereas accuracies for identifying specific species were between 88.7% and 92.7%. Transferring information from CNNs trained on large datasets (“transfer-learning”) was increasingly beneficial as the size of the training dataset decreased and raised accuracy by up to 10.3%. Removing low-confidence predictions increased model accuracies to the level of citizen scientists. By combining a trained model with classifications from citizen scientists, human effort was reduced by 43% while maintaining overall accuracy for a live experiment running on Zooniverse. Ecology researchers can significantly reduce image classification time and manual effort by combining citizen scientists and CNNs, enabling faster processing of data from large camera trap studies.

KW - animal identification

KW - camera trap

KW - citizen science

KW - convolutional neural networks

KW - deep learning

KW - machine learning

UR - http://www.scopus.com/inward/record.url?scp=85055684821&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85055684821&partnerID=8YFLogxK

U2 - 10.1111/2041-210X.13099

DO - 10.1111/2041-210X.13099

M3 - Article

VL - 10

SP - 80

EP - 91

JO - Methods in Ecology and Evolution

JF - Methods in Ecology and Evolution

SN - 2041-210X

IS - 1

ER -