Using Deep Learning to Count Albatrosses from Space: Assessing Results in Light of Ground Truth Uncertainty
Many wildlife species inhabit inaccessible environments, limiting researchers ability to conduct essential population surveys. Recently, very high resolution (sub-metre) satellite imagery has enabled remote monitoring of certain species directly from space; however, manual analysis of the imagery is...
Published in: | Remote Sensing |
---|---|
Main Authors: | , , , |
Format: | Article in Journal/Newspaper |
Language: | English |
Published: |
MDPI AG
2020
|
Subjects: | |
Online Access: | https://doi.org/10.3390/rs12122026 https://doaj.org/article/d2326768aeea48f385b21818fea46d82 |
id |
ftdoajarticles:oai:doaj.org/article:d2326768aeea48f385b21818fea46d82 |
---|---|
record_format |
openpolar |
spelling |
ftdoajarticles:oai:doaj.org/article:d2326768aeea48f385b21818fea46d82 2023-05-15T18:43:03+02:00 Using Deep Learning to Count Albatrosses from Space: Assessing Results in Light of Ground Truth Uncertainty Ellen Bowler Peter T. Fretwell Geoffrey French Michal Mackiewicz 2020-06-01T00:00:00Z https://doi.org/10.3390/rs12122026 https://doaj.org/article/d2326768aeea48f385b21818fea46d82 EN eng MDPI AG https://www.mdpi.com/2072-4292/12/12/2026 https://doaj.org/toc/2072-4292 doi:10.3390/rs12122026 2072-4292 https://doaj.org/article/d2326768aeea48f385b21818fea46d82 Remote Sensing, Vol 12, Iss 2026, p 2026 (2020) WorldView-3 convolutional neural network VHR satellite imagery wildlife monitoring observer uncertainty Wandering Albatross Science Q article 2020 ftdoajarticles https://doi.org/10.3390/rs12122026 2022-12-31T16:06:21Z Many wildlife species inhabit inaccessible environments, limiting researchers ability to conduct essential population surveys. Recently, very high resolution (sub-metre) satellite imagery has enabled remote monitoring of certain species directly from space; however, manual analysis of the imagery is time-consuming, expensive and subjective. State-of-the-art deep learning approaches can automate this process; however, often image datasets are small, and uncertainty in ground truth labels can affect supervised training schemes and the interpretation of errors. In this paper, we investigate these challenges by conducting both manual and automated counts of nesting Wandering Albatrosses on four separate islands, captured by the 31 cm resolution WorldView-3 sensor. We collect counts from six observers, and train a convolutional neural network (U-Net) using leave-one-island-out cross-validation and different combinations of ground truth labels. We show that (1) interobserver variation in manual counts is significant and differs between the four islands, (2) the small dataset can limit the networks ability to generalise to unseen imagery and (3) the choice of ground truth labels can have a significant impact on our assessment of network performance. Our final results show the network detects albatrosses as accurately as human observers for two of the islands, while in the other two misclassifications are largely caused by the presence of noise, cloud cover and habitat, which was not present in the training dataset. While the results show promise, we stress the importance of considering these factors for any study where data is limited and observer confidence is variable. Article in Journal/Newspaper Wandering Albatross Directory of Open Access Journals: DOAJ Articles Four Islands ENVELOPE(-108.218,-108.218,56.050,56.050) Remote Sensing 12 12 2026 |
institution |
Open Polar |
collection |
Directory of Open Access Journals: DOAJ Articles |
op_collection_id |
ftdoajarticles |
language |
English |
topic |
WorldView-3 convolutional neural network VHR satellite imagery wildlife monitoring observer uncertainty Wandering Albatross Science Q |
spellingShingle |
WorldView-3 convolutional neural network VHR satellite imagery wildlife monitoring observer uncertainty Wandering Albatross Science Q Ellen Bowler Peter T. Fretwell Geoffrey French Michal Mackiewicz Using Deep Learning to Count Albatrosses from Space: Assessing Results in Light of Ground Truth Uncertainty |
topic_facet |
WorldView-3 convolutional neural network VHR satellite imagery wildlife monitoring observer uncertainty Wandering Albatross Science Q |
description |
Many wildlife species inhabit inaccessible environments, limiting researchers ability to conduct essential population surveys. Recently, very high resolution (sub-metre) satellite imagery has enabled remote monitoring of certain species directly from space; however, manual analysis of the imagery is time-consuming, expensive and subjective. State-of-the-art deep learning approaches can automate this process; however, often image datasets are small, and uncertainty in ground truth labels can affect supervised training schemes and the interpretation of errors. In this paper, we investigate these challenges by conducting both manual and automated counts of nesting Wandering Albatrosses on four separate islands, captured by the 31 cm resolution WorldView-3 sensor. We collect counts from six observers, and train a convolutional neural network (U-Net) using leave-one-island-out cross-validation and different combinations of ground truth labels. We show that (1) interobserver variation in manual counts is significant and differs between the four islands, (2) the small dataset can limit the networks ability to generalise to unseen imagery and (3) the choice of ground truth labels can have a significant impact on our assessment of network performance. Our final results show the network detects albatrosses as accurately as human observers for two of the islands, while in the other two misclassifications are largely caused by the presence of noise, cloud cover and habitat, which was not present in the training dataset. While the results show promise, we stress the importance of considering these factors for any study where data is limited and observer confidence is variable. |
format |
Article in Journal/Newspaper |
author |
Ellen Bowler Peter T. Fretwell Geoffrey French Michal Mackiewicz |
author_facet |
Ellen Bowler Peter T. Fretwell Geoffrey French Michal Mackiewicz |
author_sort |
Ellen Bowler |
title |
Using Deep Learning to Count Albatrosses from Space: Assessing Results in Light of Ground Truth Uncertainty |
title_short |
Using Deep Learning to Count Albatrosses from Space: Assessing Results in Light of Ground Truth Uncertainty |
title_full |
Using Deep Learning to Count Albatrosses from Space: Assessing Results in Light of Ground Truth Uncertainty |
title_fullStr |
Using Deep Learning to Count Albatrosses from Space: Assessing Results in Light of Ground Truth Uncertainty |
title_full_unstemmed |
Using Deep Learning to Count Albatrosses from Space: Assessing Results in Light of Ground Truth Uncertainty |
title_sort |
using deep learning to count albatrosses from space: assessing results in light of ground truth uncertainty |
publisher |
MDPI AG |
publishDate |
2020 |
url |
https://doi.org/10.3390/rs12122026 https://doaj.org/article/d2326768aeea48f385b21818fea46d82 |
long_lat |
ENVELOPE(-108.218,-108.218,56.050,56.050) |
geographic |
Four Islands |
geographic_facet |
Four Islands |
genre |
Wandering Albatross |
genre_facet |
Wandering Albatross |
op_source |
Remote Sensing, Vol 12, Iss 2026, p 2026 (2020) |
op_relation |
https://www.mdpi.com/2072-4292/12/12/2026 https://doaj.org/toc/2072-4292 doi:10.3390/rs12122026 2072-4292 https://doaj.org/article/d2326768aeea48f385b21818fea46d82 |
op_doi |
https://doi.org/10.3390/rs12122026 |
container_title |
Remote Sensing |
container_volume |
12 |
container_issue |
12 |
container_start_page |
2026 |
_version_ |
1766232838239158272 |