Descripteur
Documents disponibles dans cette catégorie (7)
Ajouter le résultat dans votre panier Affiner la recherche Interroger des sources externesEtendre la recherche sur niveau(x) vers le bas
Initialization methods of convolutional neural networks for detection of image manipulations / Ivan Castillo Camacho (2021)
Titre : Initialization methods of convolutional neural networks for detection of image manipulations Titre original : Méthodes d'initialisation des réseaux de neurones convolutifs pour la détection des manipulations d'images Type de document : Thèse/HDR Auteurs : Ivan Castillo Camacho, Auteur ; Kai Wang, Directeur de thèse Editeur : Grenoble [France] : Université Grenoble Alpes Année de publication : 2021 Importance : 145 p. Format : 21 x 30 cm Note générale : bibliographie
Thèse pour obtenir le grade de Docteur de l'Université Grenoble, spécialité : signal, image, paroles, télécomsLangues : Anglais (eng) Descripteur : [Vedettes matières IGN] Traitement d'image optique
[Termes IGN] altération
[Termes IGN] apprentissage profond
[Termes IGN] classification par réseau neuronal convolutif
[Termes IGN] covariance
[Termes IGN] détection d'anomalie
[Termes IGN] estompage
[Termes IGN] filtre passe-haut
[Termes IGN] flux de données
[Termes IGN] infraction
[Termes IGN] manipulation de données
[Termes IGN] qualité des données
[Termes IGN] retouche
[Termes IGN] varianceIndex. décimale : THESE Thèses et HDR Résumé : (auteur) Fake images and videos have engulfed mass communication media. This is not something recent, manipulations and forgeries have occurred since the advent of photography itself. These alterations can go from innocent retouches in an attempt to make an image visually attractive to the spread of misleading information or even the use of false media in legal instances. Accordingly, the creation of methods that can help us assure the authenticity of an image presented as non-modified is of paramount importance. In this thesis, we aim at detecting image manipulation operations using deep learning techniques. We present three methods showing the progression of our work under one common objective, i.e, the design and test of Convolutional Neural Network (CNN) initialization methods for image forensic problems with a variance stability focus for the output of a CNN layer.First, we carry out an extensive review of the state of the art in deep-learning-based methods for image forensics. From this review we can confirm that the first layer of a CNN has big impact on the final performance. Specifically, the initialization used on the first-layer filters plays an important role that should be in line with the image forensic task in hand.As our first attempt to address this research problem, we propose a low-complexity initialization method for CNNs. Taking advantage of previous methods designed for the computer vision field, we extend the popular Xavier method to design a filter that would provide variance stability after a convolution operation. This method generates a set of random high-pass filters for the initialization of a CNN's first layer. These filters allow us to better identify forensic traces which usually lie towards the high-frequency part of the image.This first approach constitutes a good staring point of our work. However, a wrong assumption, largely utilized in the research community, was made. This is corrected in our second method where we follow a different data-dependent approach and take into consideration the real statistical properties of natural images. Accordingly, we propose a scaling method for first-layer filters which can cope well with different CNN initialization algorithms. The objective remains in keeping the stability of the variance of data flow in a CNN. We also present theoretical and experimental studies on the output variance for convolutional filter, which are the basis of our proposed data-dependent scaling.Next we describe a revisited version of our first proposal now with a corrected assumption on the statistics of natural images. More precisely, we propose an improved random high-pass initialization method which does not explicitly compute the statistics of input data. We believe that such a ``data-independent'' approach has higher flexibility and broader application range than our second method in situations where the computation of input statistics is not possible.Our proposed methods are tested over several image forensic problems and different CNN architectures.Finally, during all this thesis work we took part in a challenge competition of image forgery detection organized by the French National Research Agency and the French Directorate General of Armaments. We explain in the Appendix the objectives of the challenge along with a brief description of our work conducted for the competition. Note de contenu : 1- Introduction
2- Background knowledge and state of the art
3- Random high-pass initialization
4- Data-dependent initialization
5- Revisiting the random high-pass initialization
6- Conclusions and perspectivesNuméro de notice : 28437 Affiliation des auteurs : non IGN Thématique : IMAGERIE Nature : Thèse française Note de thèse : Thèse de Doctorat : signal, image, paroles, télécoms : Grenoble : 2021 DOI : sans En ligne : https://hal.science/tel-03346063/ Format de la ressource électronique : URL Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=98833 Forest change detection in incomplete satellite images with deep neural networks / Salman H. Khan in IEEE Transactions on geoscience and remote sensing, vol 55 n° 9 (September 2017)
Titre : Forest change detection in incomplete satellite images with deep neural networks Type de document : Article/Communication Auteurs : Salman H. Khan, Auteur ; Xuming He, Auteur ; Fatih Porikli, Auteur ; Mohammed Bennamoun, Auteur Année de publication : 2017 Article en page(s) : pp 5407 - 5423 Note générale : Bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Applications de télédétection
[Termes IGN] analyse diachronique
[Termes IGN] analyse multirésolution
[Termes IGN] apprentissage profond
[Termes IGN] détection de changement
[Termes IGN] forêt
[Termes IGN] réflectance de surface
[Termes IGN] réseau neuronal artificiel
[Termes IGN] retouche
[Termes IGN] surveillance de la végétationRésumé : (Auteur) Land cover change monitoring is an important task from the perspective of regional resource monitoring, disaster management, land development, and environmental planning. In this paper, we analyze imagery data from remote sensing satellites to detect forest cover changes over a period of 29 years (1987-2015). Since the original data are severely incomplete and contaminated with artifacts, we first devise a spatiotemporal inpainting mechanism to recover the missing surface reflectance information. The spatial filling process makes use of the available data of the nearby temporal instances followed by a sparse encoding-based reconstruction. We formulate the change detection task as a region classification problem. We build a multiresolution profile (MRP) of the target area and generate a candidate set of bounding-box proposals that enclose potential change regions. In contrast to existing methods that use handcrafted features, we automatically learn region representations using a deep neural network in a data-driven fashion. Based on these highly discriminative representations, we determine forest changes and predict their onset and offset timings by labeling the candidate set of proposals. Our approach achieves the state-of-the-art average patch classification rate of 91.6% (an improvement of ~16%) and the mean onset/offset prediction error of 4.9 months (an error reduction of five months) compared with a strong baseline. We also qualitatively analyze the detected changes in the unlabeled image regions, which demonstrate that the proposed forest change detection approach is scalable to new regions. Numéro de notice : A2017-663 Affiliation des auteurs : non IGN Thématique : FORET/IMAGERIE Nature : Article nature-HAL : ArtAvecCL-RevueIntern DOI : 10.1109/TGRS.2017.2707528 En ligne : http://dx.doi.org/10.1109/TGRS.2017.2707528 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=87105
in IEEE Transactions on geoscience and remote sensing > vol 55 n° 9 (September 2017) . - pp 5407 - 5423[article]
Titre : Disocclusion of 3D LiDAR point clouds using range images Type de document : Article/Communication Auteurs : Pierre Biasutti Année de publication : 2017 Projets : GOTMI / Papadakis, Nicolas Conférence : ISPRS 2017, Workshops HRIGI – CMRT – ISA – EuroCOW 06/06/2017 09/06/2017 Hanovre Allemagne ISPRS OA Annals Article en page(s) : pp 75 - 82 Note générale : bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Lasergrammétrie
[Termes IGN] données lidar
[Termes IGN] données localisées 3D
[Termes IGN] objet mobile
[Termes IGN] retouche
[Termes IGN] scène urbaine
[Termes IGN] semis de pointsRésumé : (auteur) This paper proposes a novel framework for the disocclusion of mobile objects in 3D LiDAR scenes aquired via street-based Mobile Mapping Systems (MMS). Most of the existing lines of research tackle this problem directly in the 3D space. This work promotes an alternative approach by using a 2D range image representation of the 3D point cloud, taking advantage of the fact that the problem of disocclusion has been intensively studied in the 2D image processing community over the past decade. First, the point cloud is turned into a 2D range image by exploiting the sensor’s topology. Using the range image, a semi-automatic segmentation procedure based on depth histograms is performed in order to select the occluding object to be removed. A variational image inpainting technique is then used to reconstruct the area occluded by that object. Finally, the range image is unprojected as a 3D point cloud. Experiments on real data prove the effectiveness of this procedure both in terms of accuracy and speed. Numéro de notice : A2017-898 Affiliation des auteurs : LASTIG MATIS+Ext (2012-2019) Thématique : IMAGERIE Nature : Article nature-HAL : ArtAvecCL-RevueIntern DOI : 10.5194/isprs-annals-IV-1-W1-75-2017 Date de publication en ligne : 30/05/2017 En ligne : https://doi.org/10.5194/isprs-annals-IV-1-W1-75-2017 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=91913
in ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences > vol IV-1/W1 (May 2017) . - pp 75 - 82[article]
Titre : Dynamic occlusion detection and inpainting of in situ captured terrestrial laser scanning point clouds sequence Type de document : Article/Communication Auteurs : Chi Chen, Auteur ; Bisheng Yang, Auteur Année de publication : 2016 Article en page(s) : pp 90 - 107 Note générale : Bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Lasergrammétrie
[Termes IGN] densité des points
[Termes IGN] détection de partie cachée
[Termes IGN] données lidar
[Termes IGN] occultation du signal
[Termes IGN] retouche
[Termes IGN] semis de points
[Termes IGN] télémétrie laser terrestre
[Termes IGN] voxelRésumé : (Auteur) Laser point clouds captured using terrestrial laser scanning (TLS) in an uncontrollable urban outdoor or indoor scene suffer from irregular shaped data blanks caused by dynamic occlusion that temporarily exists, i.e., moving objects, such as pedestrians or cars, resulting in integrality and quality losses of the scene data. This paper proposes a novel automatic dynamic occlusion detection and inpainting method for sequential TLS point clouds captured from one scan position. In situ collected laser point clouds sequences are indexed by establishing a novel panoramic space partition that assigns a three dimensional voxel to each laser point according to the scanning setups. Then two stationary background models are constructed at the ray voxel level using the laser reflectance intensity and geometrical attributes of the point set inside each voxel across the TLS sequence. Finally, the background models are combined to detect the points on the dynamic object, and the ray voxels of the detected dynamic points are tracked for further inpainting by replacing the ray voxels with the corresponding background voxels from another scan. The resulting scene is free of dynamic occlusions. Experiments validated the effectiveness of the proposed method for indoor and outdoor TLS point clouds captured by a commercial terrestrial scanner. The proposed method achieves high precision and recall rate for dynamic occlusion detection and produces clean inpainted point clouds for further processing. Numéro de notice : A2016-779 Affiliation des auteurs : non IGN Thématique : IMAGERIE Nature : Article nature-HAL : ArtAvecCL-RevueIntern DOI : 10.1016/j.isprsjprs.2016.05.007 En ligne : https://doi.org/10.1016/j.isprsjprs.2016.05.007 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=82474
in ISPRS Journal of photogrammetry and remote sensing > vol 119 (September 2016) . - pp 90 - 107[article]
Titre : Visibility estimation and joint inpainting of lidar depth maps Type de document : Article/Communication Auteurs : Marco Bevilacqua, Auteur ; Jean-François Aujol, Auteur ; Mathieu Brédif Editeur : New York : Institute of Electrical and Electronics Engineers IEEE Année de publication : 2016 Projets : 1-Pas de projet / Papadakis, Nicolas Conférence : ICIP 2016, 23rd IEEE International Conference on Image Processing 25/09/2016 28/09/2016 Phoenix Arizona - Etats-Unis Proceedings IEEE Importance : pp 3503 - 3507 Note générale : bibliographie Langues : Anglais (eng) Descripteur : [Vedettes matières IGN] Lasergrammétrie
[Termes IGN] carte de profondeur
[Termes IGN] données lidar
[Termes IGN] données localisées 3D
[Termes IGN] retouche
[Termes IGN] semis de points
[Termes IGN] visibilitéRésumé : (auteur) This paper presents a novel variational image inpainting method to solve the problem of generating, from 3-D lidar measures, a dense depth map coherent with a given color image, tackling visibility issues. When projecting the lidar point cloud onto the image plane, we generally obtain a sparse depth map, due to undersampling. Moreover, lidar and image sensor positions generally differ during acquisition, such that depth values referring to objects that are hidden from the image view point might appear with a naive projection. The proposed algorithm estimates the complete depth map, while simultaneously detecting and excluding those hidden points. It consists in a primal-dual optimization method, where a coupled total variation regularization term is included to match the depth and image gradients and a visibility indicator handles the selection of visible points. Tests with real data prove the effectiveness of the proposed strategy. Numéro de notice : C2016-035 Affiliation des auteurs : LASTIG MATIS+Ext (2012-2019) Autre URL associée : vers HAL Thématique : IMAGERIE Nature : Communication nature-HAL : ComAvecCL&ActesPubliésIntl DOI : 10.1109/ICIP.2016.7533011 Date de publication en ligne : 19/08/2016 En ligne : https://doi.org/10.1109/ICIP.2016.7533011 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=91794 Révision d'une base de connaissances : Application à la généralisation cartographique / Sylvain Bard (2000)
Permalink
