Detecting Animals in Repeated UAV Image Acquisitions by Matching CNN Activations with Optimal Transport

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

Abstract

Repeated animal censuses are crucial for wildlife parks to ensure ecological equilibriums. They are increasingly conducted using images generated by Unmanned Aerial Vehicles (UAVs), often coupled to semi-automatic object detection methods. Such methods have shown great progress also thanks to the employment of Convolutional Neural Networks (CNNs), but even the best models trained on the data acquired in one year struggle predicting animal abundances in subsequent campaigns due to the inherent shift between the datasets. In this paper we adapt a CNN-based animal detector to a follow-up UAV dataset by employing an unsupervised domain adaptation method based on Optimal Transport. We show how to infer updated labels from the source dataset by means of an ensemble of bootstraps. Our method increases the precision compared to the unmodified CNN, while not requiring additional labels from the target set.
Original languageEnglish
Title of host publication2018 IEEE International Geoscience & Remote Sensing Symposium Proceedings
Subtitle of host publicationObserving, Understanding And Forecasting The Dynamics Of Our Planet
PublisherIEEE Xplore
Pages3643-3646
ISBN (Electronic)9781538671504, 9781538671498
ISBN (Print)9781538671511
DOIs
Publication statusPublished - 5 Nov 2018
EventIGARSS 2018 - 2018 IEEE International Geoscience and Remote Sensing Symposium - Valencia, Spain
Duration: 22 Jul 201827 Jul 2018

Conference

ConferenceIGARSS 2018 - 2018 IEEE International Geoscience and Remote Sensing Symposium
CountrySpain
CityValencia
Period22/07/1827/07/18

Fingerprint

Image acquisition
Unmanned aerial vehicles (UAV)
Animals
Chemical activation
Neural networks
Labels
Detectors

Cite this

Kellenberger, B., Marcos, D., Courty, N., & Tuia, D. (2018). Detecting Animals in Repeated UAV Image Acquisitions by Matching CNN Activations with Optimal Transport. In 2018 IEEE International Geoscience & Remote Sensing Symposium Proceedings: Observing, Understanding And Forecasting The Dynamics Of Our Planet (pp. 3643-3646). IEEE Xplore. https://doi.org/10.1109/IGARSS.2018.8519012
Kellenberger, Benjamin ; Marcos, Diego ; Courty, Nicolas ; Tuia, Devis. / Detecting Animals in Repeated UAV Image Acquisitions by Matching CNN Activations with Optimal Transport. 2018 IEEE International Geoscience & Remote Sensing Symposium Proceedings: Observing, Understanding And Forecasting The Dynamics Of Our Planet. IEEE Xplore, 2018. pp. 3643-3646
@inproceedings{408a568717e54333953d6400297288e7,
title = "Detecting Animals in Repeated UAV Image Acquisitions by Matching CNN Activations with Optimal Transport",
abstract = "Repeated animal censuses are crucial for wildlife parks to ensure ecological equilibriums. They are increasingly conducted using images generated by Unmanned Aerial Vehicles (UAVs), often coupled to semi-automatic object detection methods. Such methods have shown great progress also thanks to the employment of Convolutional Neural Networks (CNNs), but even the best models trained on the data acquired in one year struggle predicting animal abundances in subsequent campaigns due to the inherent shift between the datasets. In this paper we adapt a CNN-based animal detector to a follow-up UAV dataset by employing an unsupervised domain adaptation method based on Optimal Transport. We show how to infer updated labels from the source dataset by means of an ensemble of bootstraps. Our method increases the precision compared to the unmodified CNN, while not requiring additional labels from the target set.",
author = "Benjamin Kellenberger and Diego Marcos and Nicolas Courty and Devis Tuia",
year = "2018",
month = "11",
day = "5",
doi = "10.1109/IGARSS.2018.8519012",
language = "English",
isbn = "9781538671511",
pages = "3643--3646",
booktitle = "2018 IEEE International Geoscience & Remote Sensing Symposium Proceedings",
publisher = "IEEE Xplore",

}

Kellenberger, B, Marcos, D, Courty, N & Tuia, D 2018, Detecting Animals in Repeated UAV Image Acquisitions by Matching CNN Activations with Optimal Transport. in 2018 IEEE International Geoscience & Remote Sensing Symposium Proceedings: Observing, Understanding And Forecasting The Dynamics Of Our Planet. IEEE Xplore, pp. 3643-3646, IGARSS 2018 - 2018 IEEE International Geoscience and Remote Sensing Symposium, Valencia, Spain, 22/07/18. https://doi.org/10.1109/IGARSS.2018.8519012

Detecting Animals in Repeated UAV Image Acquisitions by Matching CNN Activations with Optimal Transport. / Kellenberger, Benjamin; Marcos, Diego; Courty, Nicolas; Tuia, Devis.

2018 IEEE International Geoscience & Remote Sensing Symposium Proceedings: Observing, Understanding And Forecasting The Dynamics Of Our Planet. IEEE Xplore, 2018. p. 3643-3646.

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

TY - GEN

T1 - Detecting Animals in Repeated UAV Image Acquisitions by Matching CNN Activations with Optimal Transport

AU - Kellenberger, Benjamin

AU - Marcos, Diego

AU - Courty, Nicolas

AU - Tuia, Devis

PY - 2018/11/5

Y1 - 2018/11/5

N2 - Repeated animal censuses are crucial for wildlife parks to ensure ecological equilibriums. They are increasingly conducted using images generated by Unmanned Aerial Vehicles (UAVs), often coupled to semi-automatic object detection methods. Such methods have shown great progress also thanks to the employment of Convolutional Neural Networks (CNNs), but even the best models trained on the data acquired in one year struggle predicting animal abundances in subsequent campaigns due to the inherent shift between the datasets. In this paper we adapt a CNN-based animal detector to a follow-up UAV dataset by employing an unsupervised domain adaptation method based on Optimal Transport. We show how to infer updated labels from the source dataset by means of an ensemble of bootstraps. Our method increases the precision compared to the unmodified CNN, while not requiring additional labels from the target set.

AB - Repeated animal censuses are crucial for wildlife parks to ensure ecological equilibriums. They are increasingly conducted using images generated by Unmanned Aerial Vehicles (UAVs), often coupled to semi-automatic object detection methods. Such methods have shown great progress also thanks to the employment of Convolutional Neural Networks (CNNs), but even the best models trained on the data acquired in one year struggle predicting animal abundances in subsequent campaigns due to the inherent shift between the datasets. In this paper we adapt a CNN-based animal detector to a follow-up UAV dataset by employing an unsupervised domain adaptation method based on Optimal Transport. We show how to infer updated labels from the source dataset by means of an ensemble of bootstraps. Our method increases the precision compared to the unmodified CNN, while not requiring additional labels from the target set.

U2 - 10.1109/IGARSS.2018.8519012

DO - 10.1109/IGARSS.2018.8519012

M3 - Conference paper

SN - 9781538671511

SP - 3643

EP - 3646

BT - 2018 IEEE International Geoscience & Remote Sensing Symposium Proceedings

PB - IEEE Xplore

ER -

Kellenberger B, Marcos D, Courty N, Tuia D. Detecting Animals in Repeated UAV Image Acquisitions by Matching CNN Activations with Optimal Transport. In 2018 IEEE International Geoscience & Remote Sensing Symposium Proceedings: Observing, Understanding And Forecasting The Dynamics Of Our Planet. IEEE Xplore. 2018. p. 3643-3646 https://doi.org/10.1109/IGARSS.2018.8519012