Titre :

Spatiotemporal self-supervised pre-training on satellite imagery improves food insecurity prediction

Auteur(s) :

Cartuyvels, Ruben [Auteur]
Department of Computer Science [KU Leuven] [KU-CS]
Fierens, Tom [Auteur]
Department of Computer Science [KU Leuven] [KU-CS]
Coppieters, Emiel [Auteur]
Department of Computer Science [KU Leuven] [KU-CS]
Moens, Marie-Francine [Auteur]
Department of Computer Science [KU Leuven] [KU-CS]
Sileo, Damien [Auteur]
Inria Lille - Nord Europe
Centrale Lille
Université de Lille
Machine Learning in Information Networks [MAGNET]
Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189 [CRIStAL]

Titre de la revue :

Environmental Data Science

Pagination :

e48

Éditeur :

Cambridge University Press

Date de publication :

2023-12-18

Mot(s)-clé(s) en anglais :

deep learning food insecurity remote sensing unsupervised pre-training
deep learning
food insecurity
remote sensing
unsupervised pre-training

Discipline(s) HAL :

Informatique [cs]

Résumé en anglais : [en]

Abstract Global warming will cause unprecedented changes to the world. Predicting events such as food insecurities in specific earth regions is a valuable way to face them with adequate policies. Existing food insecurity ...
Lire la suite >Abstract Global warming will cause unprecedented changes to the world. Predicting events such as food insecurities in specific earth regions is a valuable way to face them with adequate policies. Existing food insecurity prediction models are based on handcrafted features such as population counts, food prices, or rainfall measurements. However, finding useful features is a challenging task, and data scarcity hinders accuracy. We leverage unsupervised pre-training of neural networks to automatically learn useful features from widely available L andsat -8 satellite images. We train neural feature extractors to predict whether pairs of images are coming from spatially close or distant regions on the assumption that close regions should have similar features. We also integrate a temporal dimension to our pre-training to capture the temporal trends of satellite images with improved accuracy. We show that with unsupervised pre-training on a large set of satellite images, neural feature extractors achieve a macro F1 of 65.4% on the Famine Early Warning Systems network dataset—a 24% improvement over handcrafted features. We further show that our pre-training method leads to better features than supervised learning and previous unsupervised pre-training techniques. We demonstrate the importance of the proposed time-aware pre-training and show that the pre-trained networks can predict food insecurity with limited availability of labeled data.Lire moins >

Langue :

Anglais

Vulgarisation :

Non

Collections :

• Centre de Recherche en Informatique, Signal et Automatique de Lille (CRIStAL) - UMR 9189

Source :

Harvested from HAL