Centre de documentation
scientifique

Mapping forest in the Swiss Alps treeline ecotone with explainable deep learning / Thiên-Anh Nguyen in Remote sensing of environment, vol 281 (November 2022)  

Public [article]  inRemote sensing of environment > vol 281 (November 2022) . - n° 113217 Titre : Mapping forest in the Swiss Alps treeline ecotone with explainable deep learning Type de document : Article/Communication Auteurs : Thiên-Anh Nguyen, Auteur ; Benjamin Kellenberger, Auteur ; Devis Tuia, Auteur Année de publication : 2022 Article en page(s) : n° 113217 Note générale : bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Applications de télédétection [Termes IGN] Alpes [Termes IGN] apprentissage profond [Termes IGN] canopée [Termes IGN] carte forestière [Termes IGN] classification par réseau neuronal convolutif [Termes IGN] écotone [Termes IGN] hauteur des arbres [Termes IGN] image à très haute résolution [Termes IGN] image aérienne [Termes IGN] image RVB [Termes IGN] inventaire forestier étranger (données) [Termes IGN] modèle numérique de surface de la canopée [Termes IGN] Suisse Résumé : (auteur) Forest maps are essential to understand forest dynamics. Due to the increasing availability of remote sensing data and machine learning models like convolutional neural networks, forest maps can these days be created on large scales with high accuracy. Common methods usually predict a map from remote sensing images without deliberately considering intermediate semantic concepts that are relevant to the final map. This makes the mapping process difficult to interpret, especially when using opaque deep learning models. Moreover, such procedure is entirely agnostic to the definitions of the mapping targets (e.g., forest types depending on variables such as tree height and tree density). Common models can at best learn these rules implicitly from data, which greatly hinders trust in the produced maps. In this work, we aim at building an explainable deep learning model for forest mapping that leverages prior knowledge about forest definitions to provide explanations to its decisions. We propose a model that explicitly quantifies intermediate variables like tree height and tree canopy density involved in the forest definitions, corresponding to those used to create the forest maps for training the model in the first place, and combines them accordingly. We apply our model to mapping forest types using very high resolution aerial imagery and lay particular focus on the treeline ecotone at high altitudes, where forest boundaries are complex and highly dependent on the chosen forest definition. Results show that our rule-informed model is able to quantify intermediate key variables and predict forest maps that reflect forest definitions. Through its interpretable design, it is further able to reveal implicit patterns in the manually-annotated forest labels, which facilitates the analysis of the produced maps and their comparison with other datasets. Numéro de notice : A2022-794 Affiliation des auteurs : non IGN Thématique : FORET/IMAGERIE Nature : Article nature-HAL : ArtAvecCL-RevueIntern DOI : 10.1016/j.rse.2022.113217 Date de publication en ligne : 01/09/2022 En ligne : https://doi.org/10.1016/j.rse.2022.113217 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=101928 [article]

Half a percent of labels is enough: efficient animal detection in UAV imagery using deep CNNs and active learning / Benjamin Kellenberger in IEEE Transactions on geoscience and remote sensing, vol 57 n° 12 (December 2019)  

Public [article]  inIEEE Transactions on geoscience and remote sensing > vol 57 n° 12 (December 2019) . - pp 9524 - 9533 Titre : Half a percent of labels is enough: efficient animal detection in UAV imagery using deep CNNs and active learning Type de document : Article/Communication Auteurs : Benjamin Kellenberger, Auteur ; Diego Marcos, Auteur ; Sylvain Lobry, Auteur ; Devis Tuia, Auteur Année de publication : 2019 Article en page(s) : pp 9524 - 9533 Note générale : bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Traitement d'image optique [Termes IGN] analyse d'image orientée objet [Termes IGN] apprentissage profond [Termes IGN] classification orientée objet [Termes IGN] classification par réseau neuronal [Termes IGN] détection d'objet [Termes IGN] données localisées [Termes IGN] échantillonnage de données [Termes IGN] faune locale [Termes IGN] image captée par drone [Termes IGN] Namibie [Termes IGN] objet mobile [Termes IGN] réalité de terrain [Termes IGN] recensement Résumé : (auteur) We present an Active Learning (AL) strategy for reusing a deep Convolutional Neural Network (CNN)-based object detector on a new data set. This is of particular interest for wildlife conservation: given a set of images acquired with an Unmanned Aerial Vehicle (UAV) and manually labeled ground truth, our goal is to train an animal detector that can be reused for repeated acquisitions, e.g., in follow-up years. Domain shifts between data sets typically prevent such a direct model application. We thus propose to bridge this gap using AL and introduce a new criterion called Transfer Sampling (TS). TS uses Optimal Transport (OT) to find corresponding regions between the source and the target data sets in the space of CNN activations. The CNN scores in the source data set are used to rank the samples according to their likelihood of being animals, and this ranking is transferred to the target data set. Unlike conventional AL criteria that exploit model uncertainty, TS focuses on very confident samples, thus allowing quick retrieval of true positives in the target data set, where positives are typically extremely rare and difficult to find by visual inspection. We extend TS with a new window cropping strategy that further accelerates sample retrieval. Our experiments show that with both strategies combined, less than half a percent of oracle-provided labels are enough to find almost 80% of the animals in challenging sets of UAV images, beating all baselines by a margin. Numéro de notice : A2019-598 Affiliation des auteurs : non IGN Thématique : IMAGERIE Nature : Article nature-HAL : ArtAvecCL-RevueIntern DOI : 10.1109/TGRS.2019.2927393 Date de publication en ligne : 20/08/2019 En ligne : http://doi.org/10.1109/TGRS.2019.2927393 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=94592 [article]

Land cover mapping at very high resolution with rotation equivariant CNNs : Towards small yet accurate models / Diego Marcos in ISPRS Journal of photogrammetry and remote sensing, vol 145 - part A (November 2018)  

Public [article]  inISPRS Journal of photogrammetry and remote sensing > vol 145 - part A (November 2018) Titre : Land cover mapping at very high resolution with rotation equivariant CNNs : Towards small yet accurate models Type de document : Article/Communication Auteurs : Diego Marcos, Auteur ; Michele Volpi, Auteur ; Benjamin Kellenberger, Auteur ; Devis Tuia, Auteur Année de publication : 2018 Note générale : Bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Traitement d'image optique [Termes IGN] apprentissage profond [Termes IGN] Bade-Wurtemberg (Allemagne) [Termes IGN] carte d'occupation du sol [Termes IGN] enrichissement sémantique [Termes IGN] filtrage numérique d'image [Termes IGN] image à ultra haute résolution [Termes IGN] modèle numérique de surface [Termes IGN] orthoimage [Termes IGN] réseau neuronal convolutif [Termes IGN] segmentation sémantique Résumé : (Auteur) In remote sensing images, the absolute orientation of objects is arbitrary. Depending on an object’s orientation and on a sensor’s flight path, objects of the same semantic class can be observed in different orientations in the same image. Equivariance to rotation, in this context understood as responding with a rotated semantic label map when subject to a rotation of the input image, is therefore a very desirable feature, in particular for high capacity models, such as Convolutional Neural Networks (CNNs). If rotation equivariance is encoded in the network, the model is confronted with a simpler task and does not need to learn specific (and redundant) weights to address rotated versions of the same object class. In this work we propose a CNN architecture called Rotation Equivariant Vector Field Network (RotEqNet) to encode rotation equivariance in the network itself. By using rotating convolutions as building blocks and passing only the values corresponding to the maximally activating orientation throughout the network in the form of orientation encoding vector fields, RotEqNet treats rotated versions of the same object with the same filter bank and therefore achieves state-of-the-art performances even when using very small architectures trained from scratch. We test RotEqNet in two challenging sub-decimeter resolution semantic labeling problems, and show that we can perform better than a standard CNN while requiring one order of magnitude less parameters. Numéro de notice : A2018-491 Affiliation des auteurs : non IGN Thématique : IMAGERIE/INFORMATIQUE Nature : Article nature-HAL : ArtAvecCL-RevueIntern DOI : 10.1016/j.isprsjprs.2018.01.021 Date de publication en ligne : 19/02/2018 En ligne : https://doi.org/10.1016/j.isprsjprs.2018.01.021 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=91227 [article]

Exemplaires(3)

Code-barres	Cote	Support	Localisation	Section	Disponibilité
081-2018111	RAB	Revue	Centre de documentation	En réserve L003	Disponible
081-2018113	DEP-EXM	Revue	LASTIG	Dépôt en unité	Exclu du prêt
081-2018112	DEP-EAF	Revue	Nancy	Dépôt en unité	Exclu du prêt