Descripteur
Termes IGN > informatique > intelligence artificielle > apprentissage automatique > apprentissage profond
apprentissage profond |
Documents disponibles dans cette catégorie (262)
Ajouter le résultat dans votre panier
Visionner les documents numériques
Affiner la recherche Interroger des sources externes
Etendre la recherche sur niveau(x) vers le bas
Titre : Learning to represent and reconstruct 3D deformable objects Type de document : Thèse/HDR Auteurs : Jan Bednarik, Auteur ; Pascal Fua, Directeur de thèse ; M. Salzmann, Directeur de thèse Editeur : Lausanne : Ecole Polytechnique Fédérale de Lausanne EPFL Année de publication : 2022 Importance : 138 p. Format : 21 x 30 cm Note générale : bibliographie
Thèse présentée pour l'obtention du grade de Docteur ès Sciences, Ecole Polytechnique Fédérale de LausanneLangues : Anglais (eng) Descripteur : [Vedettes matières IGN] Traitement d'image optique
[Termes IGN] appariement de formes
[Termes IGN] apprentissage profond
[Termes IGN] cohérence temporelle
[Termes IGN] déformation de surface
[Termes IGN] distorsion d'image
[Termes IGN] géométrie de Riemann
[Termes IGN] image 3D
[Termes IGN] reconstruction d'objet
[Termes IGN] semis de points
[Termes IGN] vision par ordinateurIndex. décimale : THESE Thèses et HDR Résumé : (auteur) Representing and reconstructing 3D deformable shapes are two tightly linked problems that have long been studied within the computer vision field. Deformable shapes are truly ubiquitous in the real world, whether be it specific object classes such as humans, garments and animals or more abstract ones such as generic materials deforming under stress caused by an external force. Truly practical computer vision algorithms must be able to understand the shapes of objects in the observed scenes to unlock the wide spectrum of much sought after applications ranging from virtual try-on to automated surgeries. Automatic shape reconstruction, however, is known to be an ill-posed problem, especially in the common scenario of a single image input. Therefore, the modern approaches rely on deep learning paradigm which has proven to be extremely effective even for the severely under-constrained computer vision problems. We, too, exploit the success of data-driven approaches, however, we also show that generic deep learning models can greatly benefit from being combined with explicit knowledge originating in traditional computational geometry. We analyze the use of various 3D shape representations for deformable object reconstruction and we distinctly focus on one of them, the atlas-based representation, which turns out to be especially suitable for modeling deformable shapes and which we further improve and extend to yield higher quality reconstructions. The atlas-based representation models the surfaces as an ensemble of continuous functions and thus allows for arbitrary resolution and analytical surface analysis. We identify major shortcomings of the base formulation, namely the infamous phenomena of patch collapse, patch overlap and arbitrarily strong mapping distortions, and we propose novel regularizers based on analytically computed properties of the reconstructed surfaces. Our approach counteracts the aforementioned drawbacks while yielding higher reconstruction accuracy in terms of surface normals on the tasks of single view-reconstruction, shape completion and point cloud auto-encoding. We dive into the problematics of atlas-based shape representation even deeper and focus on another pressing design flaw, the global inconsistency among the individual mappings. While the inconsistency is not reflected in the traditional reconstruction accuracy quantitative metrics, it is detrimental to the visual quality of the reconstructed surfaces. Specifically, we design loss functions encouraging intercommunication among the individual mappings which pushes the resulting surface towards a C1 smooth function. Our experiments on the tasks of single-view reconstruction and point cloud auto-encoding reveal that our method significantly improves the visual quality when compared to the baselines. Furthermore, we adapt the atlas-based representation and the related training procedure so that it could model a full sequence of a deforming object in a temporally-consistent way. In other words, the goal is to produce such reconstruction where each surface point always represents the same semantic point on the target ground-truth surface. To achieve such behavior, we note that if each surface point deforms close-to-isometrically, its semantic location likely remains unchanged. Practically, we make use of the Riemannian metric which is computed analytically on the surfaces, and force it to remain point-wise constant throughout the sequence. Our experimental results reveal that our method yields state-of-the-art results on the task of unsupervised dense shape correspondence estimation, while also improving the visual reconstruction quality. Finally, we look into a particular problem of monocular texture-less deformable shape reconstruction, an instance of the Shape-from-Shading problem. We propose a multi-task learning approach which takes an RGB image of an unknown object as the input and jointly produces a normal map, a depth map and a mesh corresponding to the observed part of the surface. We show that forcing the model to produce multiple different 3D representations of the same objects results in higher reconstruction quality. To train the network, we acquire a large real-world annotated dataset of texture-less deforming objects and we release it for public use. Finally, we prove through experiments that our approach outperforms a previous optimization based method on the single-view-reconstruction task. Note de contenu : 1- Introduction
2- Related work
3- Atlas-based representation for deformable shape reconstruction
4- Shape reconstruction by learning differentiable surface representations
5- Better patch stitching for parametric surface reconstruction
6- Temporally-consistent surface reconstruction using metrically-consistent atlases
7- Learning to reconstruct texture-less deformable surfaces from a single view
8- ConclusionNuméro de notice : 15761 Affiliation des auteurs : non IGN Thématique : IMAGERIE Nature : Thèse étrangère Note de thèse : Thèse de Doctorat : Sciences : Lausanne, EPFL : 2022 DOI : 10.5075/epfl-thesis-7974 En ligne : https://doi.org/10.5075/epfl-thesis-7974 Format de la ressource électronique : URL Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=100958 MLMT-CNN for object detection and segmentation in multi-layer and multi-spectral images / Majedaldein Almahasneh in Machine Vision and Applications, vol 33 n° 1 (January 2022)
[article]
Titre : MLMT-CNN for object detection and segmentation in multi-layer and multi-spectral images Type de document : Article/Communication Auteurs : Majedaldein Almahasneh, Auteur ; Adeline Paiement, Auteur ; Xianghua Xie, Auteur ; et al., Auteur Année de publication : 2022 Article en page(s) : n° 9 Note générale : bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Traitement d'image optique
[Termes IGN] apprentissage dirigé
[Termes IGN] apprentissage profond
[Termes IGN] atmosphère solaire
[Termes IGN] classification par réseau neuronal convolutif
[Termes IGN] couche thématique
[Termes IGN] détection d'objet
[Termes IGN] image multibande
[Termes IGN] segmentation d'imageRésumé : (auteur) Precisely localising solar Active Regions (AR) from multi-spectral images is a challenging but important task in understanding solar activity and its influence on space weather. A main challenge comes from each modality capturing a different location of the 3D objects, as opposed to typical multi-spectral imaging scenarios where all image bands observe the same scene. Thus, we refer to this special multi-spectral scenario as multi-layer. We present a multi-task deep learning framework that exploits the dependencies between image bands to produce 3D AR localisation (segmentation and detection) where different image bands (and physical locations) have their own set of results. Furthermore, to address the difficulty of producing dense AR annotations for training supervised machine learning (ML) algorithms, we adapt a training strategy based on weak labels (i.e. bounding boxes) in a recursive manner. We compare our detection and segmentation stages against baseline approaches for solar image analysis (multi-channel coronal hole detection, SPOCA for ARs) and state-of-the-art deep learning methods (Faster RCNN, U-Net). Additionally, both detection and segmentation stages are quantitatively validated on artificially created data of similar spatial configurations made from annotated multi-modal magnetic resonance images. Our framework achieves an average of 0.72 IoU (segmentation) and 0.90 F1 score (detection) across all modalities, comparing to the best performing baseline methods with scores of 0.53 and 0.58, respectively, on the artificial dataset, and 0.84 F1 score in the AR detection task comparing to baseline of 0.82 F1 score. Our segmentation results are qualitatively validated by an expert on real ARs. Numéro de notice : A2022-089 Affiliation des auteurs : non IGN Thématique : IMAGERIE Nature : Article DOI : 10.1007/s00138-021-01261-y Date de publication en ligne : 29/11/2021 En ligne : https://doi.org/10.1007/s00138-021-01261-y Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=99500
in Machine Vision and Applications > vol 33 n° 1 (January 2022) . - n° 9[article]Modeling of precipitable water vapor from GPS observations using machine learning and tomography methods / Mir Reza Ghaffari Razin in Advances in space research, vol 69 n° 7 (April 2022)
[article]
Titre : Modeling of precipitable water vapor from GPS observations using machine learning and tomography methods Type de document : Article/Communication Auteurs : Mir Reza Ghaffari Razin, Auteur ; Behzad Voosoghi, Auteur Année de publication : 2022 Article en page(s) : pp 2671 - 2681 Note générale : bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Applications de géodésie spatiale
[Termes IGN] algorithme génétique
[Termes IGN] apprentissage automatique
[Termes IGN] classification par séparateurs à vaste marge
[Termes IGN] Inférence floue
[Termes IGN] Iran
[Termes IGN] précipitation
[Termes IGN] radiosondage
[Termes IGN] réseau neuronal artificiel
[Termes IGN] retard hydrostatique
[Termes IGN] retard troposphérique zénithal
[Termes IGN] tomographie par GPS
[Termes IGN] vapeur d'eau
[Termes IGN] voxelRésumé : (auteur) This paper studies the application of two machine learning methods to model precipitable water vapor (PWV) using observations of 23 GPS stations from the local GPS network of north-west of Iran in 2011. In a first step, the zenith tropospheric delay (ZTD) and zenith hydrostatic delay (ZHD) is calculated with the Bernese GNSS software and Saastamoinen model as revised by Davis, respectively. Then, by subtracting the ZHD from the ZTD, the zenith wet delay (ZWD) is obtained at each GPS station, for all times. In a second step, ZWD is modeled by two different machine learning methods, based on the latitude, longitude, DOY, time, relative humidity, temperature and pressure. After training a Support Vector Machine (SVM) and an Artificial Neural Network (ANN), ZWD temporal and spatial variations are estimated. Using the formula by Bevis, the ZWD can be converted to PWV at any time and space, for each machine learning method. The accuracy of the two new models is evaluated using control stations, exterior and radiosonde station, whose observations were not used in the training step. Also, all the results of the SVM and ANN are compared with a voxel-based tomography (VBT) model. In the control and exterior stations, ZWD estimated by the SVM (ZWDSVM) and ANN (ZWDANN) is compared with the ZWD obtained from the GPS (ZWDGPS). Also, in the control and exterior stations, precise point positioning (PPP) is used to evaluate the accuracy of the new models. In the radiosonde station, the PWV of the new models (PWVSVM, PWVANN) is compared with the radiosonde PWV (PWVradiosonde) and voxel-based PWV (PWVVBT). The averaged relative error of the SVM, ANN and VBT models in the control stations is 10.50%, 12.71% and 12.91%, respectively. For SVM, ANN and VBT models, the averaged RMSE at the control stations is 1.87 (mm), 2.22 (mm) and 2.29 (mm), respectively. Analysis of the results of PWV estimated by the SVM, ANN and VBT, as well as the surface precipitation obtained from meteorological stations, indicate the high accuracy of the SVM in comparison with the ANN and VBT model. In the results shown in this paper, the SVM has the best ability to accurately estimate ZWD and PWV, using local GPS network observations. Numéro de notice : A2022-446 Affiliation des auteurs : non IGN Thématique : POSITIONNEMENT Nature : Article DOI : 10.1016/j.asr.2022.01.003 Date de publication en ligne : 13/01/2022 En ligne : https://doi.org/10.1016/j.asr.2022.01.003 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=100106
in Advances in space research > vol 69 n° 7 (April 2022) . - pp 2671 - 2681[article]Monitoring grassland dynamics by exploiting multi-modal satellite image time series / Anatol Garioud (2022)
Titre : Monitoring grassland dynamics by exploiting multi-modal satellite image time series Titre original : Suivi de la dynamique des prairies permanentes par analyse des séries temporelles multi-modales Type de document : Thèse/HDR Auteurs : Anatol Garioud , Auteur ; Clément Mallet , Directeur de thèse ; Silvia Valero, Directeur de thèse Editeur : Champs-sur-Marne [France] : Université Gustave Eiffel Année de publication : 2022 Importance : 194 p. Format : 21 x 30 cm Note générale : Bibliographie
Thèse présentée et soutenue en vue de l'obtention du Doctorat de l'Université Gustave Eiffel, Spécialité Sciences et Technologies de l'Information GéographiqueLangues : Anglais (eng) Descripteur : [Vedettes matières IGN] Applications de télédétection
[Termes IGN] analyse diachronique
[Termes IGN] analyse multivariée
[Termes IGN] apprentissage profond
[Termes IGN] classification par Perceptron multicouche
[Termes IGN] données auxiliaires
[Termes IGN] image Sentinel-MSI
[Termes IGN] image Sentinel-SAR
[Termes IGN] Mâcon
[Termes IGN] Normalized Difference Vegetation Index
[Termes IGN] prairie
[Termes IGN] régression
[Termes IGN] série temporelle
[Termes IGN] seuillage d'image
[Termes IGN] superpixel
[Termes IGN] surveillance agricole
[Termes IGN] ToulouseIndex. décimale : THESE Thèses et HDR Résumé : (Auteur) The vast grassland surfaces as well as the growing recognition of the ecosystem services thez provide have revealed urgent needs for their conservation and sutainable management. Despite the acknowledged importance of grassland management practices, there are currently no large-scale efforts reporting on their frequency and nature. Satellite remote sensing time series appear to be a suitable tool for efficient grassland monitoring and allow synoptic and regular analysis. The research conducted in this PhD aims to develop methods for the detection of grassland management practices from complementary optical and SAR multivariate time series. Advances in deep learning are employed to regress multivariate SAR time series and contextual knowledge towards optical NDVI. Resulting gap-free time series are used to efficiently explore methods aiming to detect vegetation status changes related to management practices on grasslands. Note de contenu : INTRODUCTION
1. Grasslands and remote sensing: context, diversity and challenges
1.1 Definition, extent and importance of grasslands
1.2 Earth observation from space: principles and applications over grasslands
1.3 Problem statement and objectives
1.4 Outline of the manuscript
2. Study areas and datasets
2.1 Study areas
2.2 Satellite data
2.3 Reference and ancillary datasets
2.4 Feature derived from sentinel images for grassland monitoring
2.5 Description of the feature engineering steps
2.6 Exploring the relationships between derived satellite features
2.7 Concluding remarks
HIGH-TEMPORAL SAMPLED TIME-SERIES
3. Sentinels regression for vegetation monitoring
3.1 Monitoring vegetation through optical-SAR synergy
3.2 Retrieving missing data in optical time series
3.3 SenRVM: a deep learning-based regression framework
3.4 Concluding remarks
4. Outcomes of the SenRVM approach
4.1 Experimental design for training and evaluating SenRVM models
4.2 Assessment of SenRVM predictions
4.3 Empirical analysis of the SenRVM results
4.4 Generalization capabilities of single-class grassland SenRVM models
4.5 Further post-processing of SenRVM results
4.6 Concluding remarks
MONITORING GRASSLANDS
5. Detecting and quantifying grassland management practices
5.1 Challenges and related work
5.2 The proposed methodology
5.3 Description of validation data
5.4 Experimental setup
5.5 Assessment of the proposed method
5.6 Potential outcomes
5.7 Concluding remarks
GENERAL CONCLUSION
6. Conclusion and perspectives
6.1 Summary
6.2 PerspectivesNuméro de notice : 26831 Affiliation des auteurs : UGE-LASTIG (2020- ) Thématique : FORET/IMAGERIE/INFORMATIQUE Nature : Thèse française Note de thèse : Thèse de Doctorat : Sciences et Technologies de l'Information Géographique : Gustave Eiffel : 2022 Organisme de stage : LASTIG (IGN) nature-HAL : Thèse DOI : sans En ligne : https://theses.hal.science/tel-03843683 Format de la ressource électronique : URL Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=100728 Réservation
Réserver ce documentExemplaires(1)
Code-barres Cote Support Localisation Section Disponibilité 26831-01 THESE Livre Centre de documentation Thèses Disponible Monitoring leaf phenology in moist tropical forests by applying a superpixel-based deep learning method to time-series images of tree canopies / Guangqin Song in ISPRS Journal of photogrammetry and remote sensing, vol 183 (January 2022)
[article]
Titre : Monitoring leaf phenology in moist tropical forests by applying a superpixel-based deep learning method to time-series images of tree canopies Type de document : Article/Communication Auteurs : Guangqin Song, Auteur ; Shengbiao Wu, Auteur ; Calvin K.F. Lee, Auteur ; et al., Auteur Année de publication : 2022 Article en page(s) : pp 19 - 33 Note générale : bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Traitement d'image optique
[Termes IGN] algorithme SLIC
[Termes IGN] apprentissage profond
[Termes IGN] canopée
[Termes IGN] classification dirigée
[Termes IGN] diagnostic foliaire
[Termes IGN] Enhanced vegetation index
[Termes IGN] feuille (végétation)
[Termes IGN] forêt tropicale
[Termes IGN] Panama
[Termes IGN] phénologie
[Termes IGN] photosynthèse
[Termes IGN] segmentation sémantique
[Termes IGN] série temporelle
[Termes IGN] superpixel
[Termes IGN] variation saisonnièreRésumé : (auteur) Tropical leaf phenology—particularly its variability at the tree-crown scale—dominates the seasonality of carbon and water fluxes. However, given enormous species diversity, accurate means of monitoring leaf phenology in tropical forests is still lacking. Time series of the Green Chromatic Coordinate (GCC) metric derived from tower-based red–greenblue (RGB) phenocams have been widely used to monitor leaf phenology in temperate forests, but its application in the tropics remains problematic. To improve monitoring of tropical phenology, we explored the use of a deep learning model (i.e. superpixel-based Residual Networks 50, SP-ResNet50) to automatically differentiate leaves from non-leaves in phenocam images and to derive leaf fraction at the tree-crown scale. To evaluate our model, we used a year of data from six phenocams in two contrasting forests in Panama. We first built a comprehensive library of leaf and non-leaf pixels across various acquisition times, exposure conditions and specific phenocams. We then divided this library into training and testing components. We evaluated the model at three levels: 1) superpixel level with a testing set, 2) crown level by comparing the model-derived leaf fractions with those derived using image-specific supervised classification, and 3) temporally using all daily images to assess the diurnal stability of the model-derived leaf fraction. Finally, we compared the model-derived leaf fraction phenology with leaf phenology derived from GCC. Our results show that: 1) the SP-ResNet50 model accurately differentiates leaves from non-leaves (overall accuracy of 93%) and is robust across all three levels of evaluations; 2) the model accurately quantifies leaf fraction phenology across tree-crowns and forest ecosystems; and 3) the combined use of leaf fraction and GCC helps infer the timing of leaf emergence, maturation and senescence, critical information for modeling photosynthetic seasonality of tropical forests. Collectively, this study offers an improved means for automated tropical phenology monitoring using phenocams. Numéro de notice : A2022-009 Affiliation des auteurs : non IGN Thématique : FORET/IMAGERIE Nature : Article nature-HAL : ArtAvecCL-RevueIntern DOI : 10.1016/j.isprsjprs.2021.10.023 Date de publication en ligne : 10/11/2021 En ligne : https://doi.org/10.1016/j.isprsjprs.2021.10.023 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=99057
in ISPRS Journal of photogrammetry and remote sensing > vol 183 (January 2022) . - pp 19 - 33[article]Réservation
Réserver ce documentExemplaires(3)
Code-barres Cote Support Localisation Section Disponibilité 081-2022011 SL Revue Centre de documentation Revues en salle Disponible 081-2022013 DEP-RECP Revue LASTIG Dépôt en unité Exclu du prêt 081-2022012 DEP-RECF Revue Nancy Dépôt en unité Exclu du prêt PermalinkOptimization of deep neural networks: A functional perspective with applications in image classification / Simon Roburin (2022)PermalinkPhotogrammetric point clouds: quality assessment, filtering, and change detection / Zhenchao Zhang (2022)PermalinkReprésentation et combinaison de l'information géographique pour l'apprentissage profond / Azelle Courtial (2022)PermalinkRepresenting vector geographic information as a tensor for deep learning based map generalisation / Azelle Courtial (2022)PermalinkPermalinkPermalinkSelf-attention and generative adversarial networks for algae monitoring / Nhut Hai Huynh in European journal of remote sensing, vol 55 n° 1 (2022)PermalinkSemantic segmentation of high-resolution remote sensing images based on a class feature attention mechanism fused with Deeplabv3+ / Zhimin Wang in Computers & geosciences, vol 158 (January 2022)PermalinkStudying informativeness of satellite image texture for sea ice state retrieval using deep learning methods / Clément Fougerouse (2022)Permalink