Exploring data quantity requirements for Domain Adaptation in the classification of satellite image time series

Benjamin Lucas, Charlotte Pelletier, Jordi Inglada, Daniel Schmidt, Geoffrey I. Webb, Francois Petitjean

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

Abstract

Land cover maps are a vital input variable in all types of environmental research and management. However the modern state-of-The-Art machine learning techniques used to create them require substantial training data to produce optimal accuracy. Domain Adaptation is one technique researchers might use when labelled training data are unavailable or scarce. This paper looks at the result of training a convolutional neural network model on a region where data are available (source domain), and then adapting this model to another region (target domain) by retraining it on the available labelled data, and in particular how these results change with increasing data availability. Our experiments performing domain adaptation on satellite image time series, draw three interesting conclusions: (1) a model trained only on data from the source domain delivers 73.0% test accuracy on the target domain; (2) when all of the weights are retrained on the target data, over 16,000 instances were required to improve upon the accuracy of the source-only model; and (3) even if sufficient data is available in the target domain, using a model pretrained on a source domain will result in better overall test accuracy compared to a model trained on target domain data only-88.9% versus 84.7%.

Original languageEnglish
Title of host publicationMultiTemp 2019, 10th International Workshop on the Analysis of Multitemporal Remote Sensing Images, August 5-7, 2019 – Shanghai, China
EditorsFrancesca Bovolo, Sicong Liu
Place of PublicationPiscataway NJ USA
PublisherIEEE, Institute of Electrical and Electronics Engineers
Number of pages4
ISBN (Electronic)9781728146157
ISBN (Print)9781728146164
DOIs
Publication statusPublished - 2019
EventInternational Workshop on the Analysis of Multitemporal Remote Sensing Images 2019 - Shanghai, China
Duration: 5 Aug 20197 Aug 2019
Conference number: 10th
https://multitemp2019.tongji.edu.cn/

Conference

ConferenceInternational Workshop on the Analysis of Multitemporal Remote Sensing Images 2019
Abbreviated titleMultiTemp 2019
CountryChina
CityShanghai
Period5/08/197/08/19
Internet address

Cite this

Lucas, B., Pelletier, C., Inglada, J., Schmidt, D., Webb, G. I., & Petitjean, F. (2019). Exploring data quantity requirements for Domain Adaptation in the classification of satellite image time series. In F. Bovolo, & S. Liu (Eds.), MultiTemp 2019, 10th International Workshop on the Analysis of Multitemporal Remote Sensing Images, August 5-7, 2019 – Shanghai, China [8866898] Piscataway NJ USA: IEEE, Institute of Electrical and Electronics Engineers. https://doi.org/10.1109/Multi-Temp.2019.8866898
Lucas, Benjamin ; Pelletier, Charlotte ; Inglada, Jordi ; Schmidt, Daniel ; Webb, Geoffrey I. ; Petitjean, Francois. / Exploring data quantity requirements for Domain Adaptation in the classification of satellite image time series. MultiTemp 2019, 10th International Workshop on the Analysis of Multitemporal Remote Sensing Images, August 5-7, 2019 – Shanghai, China. editor / Francesca Bovolo ; Sicong Liu. Piscataway NJ USA : IEEE, Institute of Electrical and Electronics Engineers, 2019.
@inproceedings{67b84356920c4a8183d89e5303dbd442,
title = "Exploring data quantity requirements for Domain Adaptation in the classification of satellite image time series",
abstract = "Land cover maps are a vital input variable in all types of environmental research and management. However the modern state-of-The-Art machine learning techniques used to create them require substantial training data to produce optimal accuracy. Domain Adaptation is one technique researchers might use when labelled training data are unavailable or scarce. This paper looks at the result of training a convolutional neural network model on a region where data are available (source domain), and then adapting this model to another region (target domain) by retraining it on the available labelled data, and in particular how these results change with increasing data availability. Our experiments performing domain adaptation on satellite image time series, draw three interesting conclusions: (1) a model trained only on data from the source domain delivers 73.0{\%} test accuracy on the target domain; (2) when all of the weights are retrained on the target data, over 16,000 instances were required to improve upon the accuracy of the source-only model; and (3) even if sufficient data is available in the target domain, using a model pretrained on a source domain will result in better overall test accuracy compared to a model trained on target domain data only-88.9{\%} versus 84.7{\%}.",
author = "Benjamin Lucas and Charlotte Pelletier and Jordi Inglada and Daniel Schmidt and Webb, {Geoffrey I.} and Francois Petitjean",
year = "2019",
doi = "10.1109/Multi-Temp.2019.8866898",
language = "English",
isbn = "9781728146164",
editor = "Bovolo, {Francesca } and Liu, {Sicong }",
booktitle = "MultiTemp 2019, 10th International Workshop on the Analysis of Multitemporal Remote Sensing Images, August 5-7, 2019 – Shanghai, China",
publisher = "IEEE, Institute of Electrical and Electronics Engineers",
address = "United States of America",

}

Lucas, B, Pelletier, C, Inglada, J, Schmidt, D, Webb, GI & Petitjean, F 2019, Exploring data quantity requirements for Domain Adaptation in the classification of satellite image time series. in F Bovolo & S Liu (eds), MultiTemp 2019, 10th International Workshop on the Analysis of Multitemporal Remote Sensing Images, August 5-7, 2019 – Shanghai, China., 8866898, IEEE, Institute of Electrical and Electronics Engineers, Piscataway NJ USA, International Workshop on the Analysis of Multitemporal Remote Sensing Images 2019, Shanghai, China, 5/08/19. https://doi.org/10.1109/Multi-Temp.2019.8866898

Exploring data quantity requirements for Domain Adaptation in the classification of satellite image time series. / Lucas, Benjamin; Pelletier, Charlotte; Inglada, Jordi; Schmidt, Daniel; Webb, Geoffrey I.; Petitjean, Francois.

MultiTemp 2019, 10th International Workshop on the Analysis of Multitemporal Remote Sensing Images, August 5-7, 2019 – Shanghai, China. ed. / Francesca Bovolo; Sicong Liu. Piscataway NJ USA : IEEE, Institute of Electrical and Electronics Engineers, 2019. 8866898.

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

TY - GEN

T1 - Exploring data quantity requirements for Domain Adaptation in the classification of satellite image time series

AU - Lucas, Benjamin

AU - Pelletier, Charlotte

AU - Inglada, Jordi

AU - Schmidt, Daniel

AU - Webb, Geoffrey I.

AU - Petitjean, Francois

PY - 2019

Y1 - 2019

N2 - Land cover maps are a vital input variable in all types of environmental research and management. However the modern state-of-The-Art machine learning techniques used to create them require substantial training data to produce optimal accuracy. Domain Adaptation is one technique researchers might use when labelled training data are unavailable or scarce. This paper looks at the result of training a convolutional neural network model on a region where data are available (source domain), and then adapting this model to another region (target domain) by retraining it on the available labelled data, and in particular how these results change with increasing data availability. Our experiments performing domain adaptation on satellite image time series, draw three interesting conclusions: (1) a model trained only on data from the source domain delivers 73.0% test accuracy on the target domain; (2) when all of the weights are retrained on the target data, over 16,000 instances were required to improve upon the accuracy of the source-only model; and (3) even if sufficient data is available in the target domain, using a model pretrained on a source domain will result in better overall test accuracy compared to a model trained on target domain data only-88.9% versus 84.7%.

AB - Land cover maps are a vital input variable in all types of environmental research and management. However the modern state-of-The-Art machine learning techniques used to create them require substantial training data to produce optimal accuracy. Domain Adaptation is one technique researchers might use when labelled training data are unavailable or scarce. This paper looks at the result of training a convolutional neural network model on a region where data are available (source domain), and then adapting this model to another region (target domain) by retraining it on the available labelled data, and in particular how these results change with increasing data availability. Our experiments performing domain adaptation on satellite image time series, draw three interesting conclusions: (1) a model trained only on data from the source domain delivers 73.0% test accuracy on the target domain; (2) when all of the weights are retrained on the target data, over 16,000 instances were required to improve upon the accuracy of the source-only model; and (3) even if sufficient data is available in the target domain, using a model pretrained on a source domain will result in better overall test accuracy compared to a model trained on target domain data only-88.9% versus 84.7%.

UR - http://www.scopus.com/inward/record.url?scp=85074298390&partnerID=8YFLogxK

U2 - 10.1109/Multi-Temp.2019.8866898

DO - 10.1109/Multi-Temp.2019.8866898

M3 - Conference Paper

SN - 9781728146164

BT - MultiTemp 2019, 10th International Workshop on the Analysis of Multitemporal Remote Sensing Images, August 5-7, 2019 – Shanghai, China

A2 - Bovolo, Francesca

A2 - Liu, Sicong

PB - IEEE, Institute of Electrical and Electronics Engineers

CY - Piscataway NJ USA

ER -

Lucas B, Pelletier C, Inglada J, Schmidt D, Webb GI, Petitjean F. Exploring data quantity requirements for Domain Adaptation in the classification of satellite image time series. In Bovolo F, Liu S, editors, MultiTemp 2019, 10th International Workshop on the Analysis of Multitemporal Remote Sensing Images, August 5-7, 2019 – Shanghai, China. Piscataway NJ USA: IEEE, Institute of Electrical and Electronics Engineers. 2019. 8866898 https://doi.org/10.1109/Multi-Temp.2019.8866898