{"id":6792,"date":"2020-08-12T15:10:28","date_gmt":"2020-08-12T14:10:28","guid":{"rendered":"https:\/\/datascientest.com\/?p=6792"},"modified":"2023-11-09T14:12:45","modified_gmt":"2023-11-09T14:12:45","slug":"deep-learning-et-style-transfer","status":"publish","type":"post","link":"https:\/\/datascientest.com\/deep-learning-et-style-transfer","title":{"rendered":"Deep learning &amp; Style transfer"},"content":{"rendered":"\t\t<div data-elementor-type=\"wp-post\" data-elementor-id=\"6792\" class=\"elementor elementor-6792\" data-elementor-post-type=\"post\">\n\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-a1d7878 e-flex e-con-boxed e-con\" data-id=\"a1d7878\" data-element_type=\"container\" data-settings=\"{&quot;content_width&quot;:&quot;boxed&quot;}\">\n\t\t\t\t\t<div class=\"e-con-inner\">\n\t\t\t\t<div class=\"elementor-element elementor-element-1642f01 elementor-widget elementor-widget-heading\" data-id=\"1642f01\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<style>\/*! elementor - v3.15.0 - 02-08-2023 *\/\n.elementor-heading-title{padding:0;margin:0;line-height:1}.elementor-widget-heading .elementor-heading-title[class*=elementor-size-]>a{color:inherit;font-size:inherit;line-height:inherit}.elementor-widget-heading .elementor-heading-title.elementor-size-small{font-size:15px}.elementor-widget-heading .elementor-heading-title.elementor-size-medium{font-size:19px}.elementor-widget-heading .elementor-heading-title.elementor-size-large{font-size:29px}.elementor-widget-heading .elementor-heading-title.elementor-size-xl{font-size:39px}.elementor-widget-heading .elementor-heading-title.elementor-size-xxl{font-size:59px}<\/style><h2 class=\"elementor-heading-title elementor-size-default\"><p>Avez-vous d\u00e9j\u00e0 souhait\u00e9 pouvoir peindre comme Van Gogh? Reproduire l\u2019esth\u00e9tique de Monet dans des paysages modernes?<\/p>\n<p>Dans cet article, nous allons vous pr\u00e9senter une technique utilisant le Deep Learning pour appliquer \u00e0 une image d\u2019origine le style d\u2019une autre. Cette technique d\u2019optimisation est connue sous le nom de neural style transfer (ou transfert de style), d\u00e9crite pour la premi\u00e8re fois dans l\u2019article de Leon A. Gatys, <a href\"https:\/\/arxiv.org\/abs\/1508.06576\">A Neural Algorithm of Artistic Style<\/a>.<\/p><\/h2>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-7214bc1 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"7214bc1\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-53d88e1\" data-id=\"53d88e1\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-24a9c06 elementor-widget elementor-widget-heading\" data-id=\"24a9c06\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h3 class=\"elementor-heading-title elementor-size-default\">Qu\u2019est ce que le style transfer ?<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-0231c00 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"0231c00\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-8e724d7\" data-id=\"8e724d7\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-e475194 elementor-widget elementor-widget-text-editor\" data-id=\"e475194\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<style>\/*! elementor - v3.15.0 - 02-08-2023 *\/\n.elementor-widget-text-editor.elementor-drop-cap-view-stacked .elementor-drop-cap{background-color:#69727d;color:#fff}.elementor-widget-text-editor.elementor-drop-cap-view-framed .elementor-drop-cap{color:#69727d;border:3px solid;background-color:transparent}.elementor-widget-text-editor:not(.elementor-drop-cap-view-default) .elementor-drop-cap{margin-top:8px}.elementor-widget-text-editor:not(.elementor-drop-cap-view-default) .elementor-drop-cap-letter{width:1em;height:1em}.elementor-widget-text-editor .elementor-drop-cap{float:left;text-align:center;line-height:1;font-size:50px}.elementor-widget-text-editor .elementor-drop-cap-letter{display:inline-block}<\/style>\t\t\t\t<p><span style=\"font-weight: 400;\">Le <\/span><i><span style=\"font-weight: 400;\">style transfer<\/span><\/i><span style=\"font-weight: 400;\"> est une des applications les plus cr\u00e9atives des R\u00e9seaux de Neurones Convolutifs (pour en savoir plus sur les CNN, nous avons publi\u00e9 r\u00e9cemment <a href=\"https:\/\/datascientest.com\/convolutional-neural-network\">un article sur le sujet<\/a>)<\/span><span style=\"font-weight: 400;\">.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">Elle permet de <\/span><b>r\u00e9cup\u00e9rer le style d\u2019une image<\/b><span style=\"font-weight: 400;\"> et de <\/span><b>le restituer \u00e0 n\u2019importe quelle autre image<\/b><span style=\"font-weight: 400;\">. C\u2019est une technique int\u00e9ressante qui permet de mettre en \u00e9vidence les capacit\u00e9s et les repr\u00e9sentations internes des r\u00e9seaux de neurones. Elle peut s\u2019av\u00e9rer \u00e9galement utile dans certains domaines scientifiques pour l\u2019augmentation ou la simulation de donn\u00e9es images.<\/span><\/p><p><span style=\"font-weight: 400;\">Les combinaisons presque infinies de contenus et de styles possibles font \u00e9merger des r\u00e9sultats uniques et toujours plus cr\u00e9atifs chez les adeptes de r\u00e9seaux de neurones, et parfois m\u00eame de vrais chefs-d&rsquo;\u0153uvre<\/span><b>.<\/b><\/p><blockquote><p><span style=\"font-weight: 400;\">Pour fonctionner, il est n\u00e9cessaire de choisir une <\/span><b>image de r\u00e9f\u00e9rence pour le contenu<\/b><span style=\"font-weight: 400;\"> et d\u2019une<\/span><b> image de r\u00e9f\u00e9rence pour le style <\/b><span style=\"font-weight: 400;\">(comme l\u2019\u0153uvre d&rsquo;art d&rsquo;un peintre c\u00e9l\u00e8bre). Une <\/span><b>troisi\u00e8me image<\/b><span style=\"font-weight: 400;\">, initialis\u00e9e avec l\u2019image de contenu est ensuite <\/span><b>optimis\u00e9e petit \u00e0 petit<\/b><span style=\"font-weight: 400;\">, jusqu\u2019\u00e0 ce qu\u2019elle ressemble \u00e0 l&rsquo;image de contenu, mais \u00ab\u00a0peinte\u00a0\u00bb dans le style de la seconde image.<\/span><\/p><\/blockquote><p><span style=\"font-weight: 400;\">Le principe du transfert de style consiste principalement \u00e0 <\/span><b>d\u00e9finir deux fonctions de distance<\/b><span style=\"font-weight: 400;\">:<\/span><\/p><ul><li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">l&rsquo;une qui d\u00e9crit la diff\u00e9rence entre le <\/span><b>contenu<\/b><span style=\"font-weight: 400;\"> de deux images<\/span><\/li><li style=\"font-weight: 400;\"><span style=\"font-weight: 400;\">l&rsquo;autre qui d\u00e9crit la diff\u00e9rence entre deux images par rapport \u00e0 leurs <\/span><b>styles<\/b><\/li><\/ul><p><span style=\"font-weight: 400;\">Il suffit ensuite de <\/span><b>minimiser ces deux distances<\/b><span style=\"font-weight: 400;\">, en utilisant une technique de r\u00e9tropropagation pour obtenir apr\u00e8s optimisation, une image qui correspond au contenu de l\u2019image de contenu et au style de l\u2019image de style.<\/span><\/p><p><span style=\"font-weight: 400;\">Les distances que nous utiliserons seront d\u00e9crites plus bas et seront calcul\u00e9es \u00e0 partir des images extraites de couches interm\u00e9diaires du r\u00e9seau de neurones.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">Un aspect impressionnant de cette technique est qu&rsquo;aucun nouvel entra\u00eenement de r\u00e9seaux de neurones\u00a0 n&rsquo;est n\u00e9cessaire &#8211; l\u2019utilisation de r\u00e9seaux pr\u00e9-entrain\u00e9s comme VGG19 suffit et fonctionne amplement. Vous vous demandez comment cela fonctionne, n\u2019h\u00e9sitez pas \u00e0 consulter notre article sur le <a href=\"https:\/\/datascientest.com\/deep_learning#transferlearning\">Transfer Learning<\/a><\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-6bd636a elementor-widget elementor-widget-text-editor\" data-id=\"6bd636a\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><b>Dans cet article je vous pr\u00e9senterai un exemple concret d\u2019utilisation du transfert de style en utilisant la biblioth\u00e8que <\/b><a href=\"https:\/\/www.tensorflow.org\/?hl=fr\">Tensorflow<\/a><\/p><p><span style=\"font-weight: 400;\">La d\u00e9marche utilis\u00e9e ici imite le plus fid\u00e8lement possible les m\u00e9thodes du papier originale. Le code complet de cet article est disponible \u00e0 l&rsquo;<a href=\"https:\/\/github.com\/DataScientest\/blog\/blob\/master\/style_transfer\/style_transfer.ipynb\">adresse suivante<\/a>.<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-27e759a elementor-widget elementor-widget-text-editor\" data-id=\"27e759a\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">En utilisant seulement deux images de base \u00e0 la fois, nous pourrons cr\u00e9er des chefs-d\u2019\u0153uvre comme cette repr\u00e9sentation de\u00a0 <\/span><i><span style=\"font-weight: 400;\">La Naissance de V\u00e9nus<\/span><\/i><span style=\"font-weight: 400;\"> de Botticelli, si elle avait \u00e9t\u00e9 peinte par Marc Chagall 4 si\u00e8cles plus tard.<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-a101450 elementor-widget elementor-widget-image\" data-id=\"a101450\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<style>\/*! elementor - v3.15.0 - 02-08-2023 *\/\n.elementor-widget-image{text-align:center}.elementor-widget-image a{display:inline-block}.elementor-widget-image a img[src$=\".svg\"]{width:48px}.elementor-widget-image img{vertical-align:middle;display:inline-block}<\/style>\t\t\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"601\" height=\"395\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/3.png\" class=\"attachment-large size-large wp-image-6962\" alt=\"\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/3.png 601w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/3-300x197.png 300w\" sizes=\"(max-width: 601px) 100vw, 601px\" \/>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-a38fa62 elementor-widget elementor-widget-text-editor\" data-id=\"a38fa62\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Pour notre exemple, nous avons besoin de deux images de base que nous voulons \u00ab\u00a0m\u00e9langer\u00a0\u00bb ensemble. La premi\u00e8re repr\u00e9sente le contenu que nous souhaitons conserver. Dans mon cas, j&rsquo;utiliserai cette peinture bien connue de <\/span><i><span style=\"font-weight: 400;\">La Naissance de V\u00e9nus <\/span><\/i><span style=\"font-weight: 400;\">de Sandro Botticelli \u00a0 :<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-847e69a elementor-widget elementor-widget-image\" data-id=\"847e69a\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<a href=\"https:\/\/www.kazoart.com\/blog\/loeuvre-a-la-loupe-la-naissance-de-venus-botticelli\/\">\n\t\t\t\t\t\t\t<img decoding=\"async\" width=\"800\" height=\"502\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/2-1-1024x643.jpg\" class=\"attachment-large size-large wp-image-6983\" alt=\"\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/2-1-1024x643.jpg 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/2-1-300x188.jpg 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/2-1-768x482.jpg 768w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/2-1.jpg 1280w\" sizes=\"(max-width: 800px) 100vw, 800px\" \/>\t\t\t\t\t\t\t\t<\/a>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-cb7efbd elementor-widget elementor-widget-text-editor\" data-id=\"cb7efbd\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">La seconde image contiendra le style que nous voulons garder et appliquer \u00e0 la premi\u00e8re.<\/span><\/p><p><span style=\"font-weight: 400;\">Ici, je choisis le style incomparable du peintre embl\u00e9matique du 20\u00e8me si\u00e8cle, Marc Chagall, gr\u00e2ce \u00e0 l\u2019une de ses \u0153uvres intitul\u00e9e <\/span><i><span style=\"font-weight: 400;\">La vie<\/span><\/i><span style=\"font-weight: 400;\">.<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-8d0fbf8 elementor-widget elementor-widget-image\" data-id=\"8d0fbf8\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<a href=\"https:\/\/i.pinimg.com\/originals\/6f\/dc\/7b\/6fdc7b6ca8cd57e42dc09d46a23e12b1.jpg\">\n\t\t\t\t\t\t\t<img decoding=\"async\" width=\"800\" height=\"608\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/1-2-1024x778.jpg\" class=\"attachment-large size-large wp-image-6984\" alt=\"\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/1-2-1024x778.jpg 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/1-2-300x228.jpg 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/1-2-768x584.jpg 768w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/1-2.jpg 1397w\" sizes=\"(max-width: 800px) 100vw, 800px\" \/>\t\t\t\t\t\t\t\t<\/a>\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-71cdc09 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"71cdc09\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-a0cc0ce\" data-id=\"a0cc0ce\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-8b48c8b elementor-widget elementor-widget-heading\" data-id=\"8b48c8b\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h3 class=\"elementor-heading-title elementor-size-default\">Chargement et pre-processing des images\n<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-96b8484 elementor-widget elementor-widget-text-editor\" data-id=\"96b8484\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Nous allons dans un premier temps<\/span><b> charger et pr\u00e9parer les donn\u00e9es <\/b><span style=\"font-weight: 400;\">pour leur appliquer\u00a0 le m\u00eame processus de pr\u00e9traitement que celui utilis\u00e9 pour l\u2019entra\u00eenement des r\u00e9seaux VGG. Puis nous allons \u00e9galement cr\u00e9er une fonction qui effectue le processing inverse pour pouvoir afficher notre image finale.<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-ec68793 elementor-widget elementor-widget-shortcode\" data-id=\"ec68793\" data-element_type=\"widget\" data-widget_type=\"shortcode.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div class=\"elementor-shortcode\">[<script src=\"https:\/\/gist.github.com\/Raphkas15\/613fcc6c055895fe4ca383ad91ce3e6e.js\"><\/script>]\n<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-e497d2b elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"e497d2b\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-2e0bd8c\" data-id=\"2e0bd8c\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-846c302 elementor-widget elementor-widget-heading\" data-id=\"846c302\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h4 class=\"elementor-heading-title elementor-size-default\">Comment discerner le style et le contenu d\u2019une image ?<\/h4>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-feca150 elementor-widget elementor-widget-text-editor\" data-id=\"feca150\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Lorsqu\u2019un r\u00e9seau de neurones convolutif (comme VGG19) est entra\u00een\u00e9 pour effectuer une classification d\u2019images, il doit comprendre l&rsquo;image.<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-e23cff3 elementor-widget elementor-widget-text-editor\" data-id=\"e23cff3\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><b>\u00c0 l\u2019inverse du Machine Learning \u2018classique\u2019, en Deep Learning, les <\/b><b><i>features <\/i><\/b><b>sont g\u00e9n\u00e9r\u00e9 par le r\u00e9seau lui-m\u00eame pendant l\u2019apprentissage par <\/b><b><i>r\u00e9tropropagation<\/i><\/b><b>. <\/b><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-db13f58 elementor-widget elementor-widget-text-editor\" data-id=\"db13f58\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<blockquote><p><span style=\"font-weight: 400;\">Lors de l\u2019entra\u00eenement, le r\u00e9seau de neurones se sert des pixels des images fournies afin d\u2019en construire une repr\u00e9sentation interne par le biais de transformations diverses.\u00a0 Une fois le r\u00e9seau entra\u00een\u00e9, l&rsquo;image brute en entr\u00e9e est transform\u00e9e\u00a0 en une succession de caract\u00e9ristiques complexes pr\u00e9sentes dans l&rsquo;image afin de pr\u00e9dire la classe \u00e0 laquelle elle appartient. <\/span><\/p><\/blockquote>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-d85199c elementor-widget elementor-widget-text-editor\" data-id=\"d85199c\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Ainsi, dans les couches interm\u00e9diaires du r\u00e9seau, il est possible d\u2019extraire certaines des caract\u00e9ristiques d\u2019une image afin de d\u00e9crire son contenu et son style.<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-23a0364 elementor-widget elementor-widget-text-editor\" data-id=\"23a0364\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Afin d&rsquo;obtenir \u00e0 la fois le contenu et le style de nos images, nous allons donc <\/span><b>extraire certaines couches interm\u00e9diaires<\/b><span style=\"font-weight: 400;\"> de notre mod\u00e8le.\u00a0 En partant de la couche d&rsquo;entr\u00e9e du r\u00e9seau, les premi\u00e8res activations de couche repr\u00e9sentent des caract\u00e9ristiques de bas niveau comme les bords et les textures. Les derni\u00e8res couches repr\u00e9sentent les caract\u00e9ristiques de plus haut niveau, comme les oreilles ou les yeux.<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-e22c5c3 elementor-widget elementor-widget-text-editor\" data-id=\"e22c5c3\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Nous utiliserons donc une des derni\u00e8res couches pour repr\u00e9senter le contenu, et un ensemble de 5 couches (les premi\u00e8res de chaque bloc) pour repr\u00e9senter le style des images.<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-b553ba4 elementor-widget elementor-widget-shortcode\" data-id=\"b553ba4\" data-element_type=\"widget\" data-widget_type=\"shortcode.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div class=\"elementor-shortcode\"><script src=\"https:\/\/gist.github.com\/Raphkas15\/cba86a01f36be4a61d92745a937144d8.js\"><\/script><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-541e6e8 elementor-widget elementor-widget-text-editor\" data-id=\"541e6e8\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Nous chargeons le <\/span><b>mod\u00e8le pr\u00e9-entra\u00een\u00e9 avec ses poids<\/b><span style=\"font-weight: 400;\"> et cr\u00e9ons un nouveau mod\u00e8le qui prend en entr\u00e9e une image et<\/span><b> retourne les couches interm\u00e9diaires<\/b><span style=\"font-weight: 400;\"> correspondant au <\/span><b>contenu <\/b><span style=\"font-weight: 400;\">et au <\/span><b>style<\/b><span style=\"font-weight: 400;\"> de l\u2019image.<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-ac2bac9 elementor-widget elementor-widget-shortcode\" data-id=\"ac2bac9\" data-element_type=\"widget\" data-widget_type=\"shortcode.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div class=\"elementor-shortcode\"><script src=\"https:\/\/gist.github.com\/Raphkas15\/eeddd94559c7095aac9e00a5842a897c.js\"><\/script><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-e7e3df9 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"e7e3df9\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-0454326\" data-id=\"0454326\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-36ad919 elementor-widget elementor-widget-heading\" data-id=\"36ad919\" data-element_type=\"widget\" data-widget_type=\"heading.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<h3 class=\"elementor-heading-title elementor-size-default\">D\u00e9finition et cr\u00e9ation des fonctions de perte<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-9b5aab0 elementor-widget elementor-widget-text-editor\" data-id=\"9b5aab0\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Notre d\u00e9finition de la fonction de perte pour le\u00a0 contenu est assez simple. <\/span><b>Nous transmettons au r\u00e9seau \u00e0 la fois l&rsquo;image du contenu souhait\u00e9 et notre image d&rsquo;entr\u00e9e de base<\/b><span style=\"font-weight: 400;\">. Cela permettra de renvoyer les sorties des couches interm\u00e9diaires\u00a0 de notre mod\u00e8le. Ensuite, nous prenons simplement <\/span><b>l\u2019erreur quadratique moyenne entre les deux repr\u00e9sentations interm\u00e9diaires<\/b><span style=\"font-weight: 400;\"> de ces images.<\/span><\/p><blockquote><p><span style=\"font-weight: 400;\">Le but de cette fonction de perte est de s&rsquo;assurer que l&rsquo;image g\u00e9n\u00e9r\u00e9e garde certaines des caract\u00e9ristiques \u00ab\u00a0globales\u00a0\u00bb de l&rsquo;image de contenu.\u00a0<\/span><\/p><\/blockquote><p><span style=\"font-weight: 400;\">Dans notre exemple, nous voulons nous assurer que l&rsquo;image g\u00e9n\u00e9r\u00e9e restitue la pose de la d\u00e9esse debout dans la conque d\u2019un coquillage, ainsi que les personnages qui l\u2019entourent. Cela signifie que des formes telles que les corps, les cheveux, et le coquillage doivent \u00eatre reconnaissables.<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-329af04 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"329af04\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-1ba9161\" data-id=\"1ba9161\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-a04ad7f elementor-widget elementor-widget-text-editor\" data-id=\"a04ad7f\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Nous effectuons la r\u00e9tropropagation de la mani\u00e8re habituelle de mani\u00e8re \u00e0 minimiser cette perte de contenu. <\/span><b>Nous modifions donc l&rsquo;image initiale jusqu&rsquo;\u00e0 ce qu&rsquo;elle g\u00e9n\u00e8re une r\u00e9ponse similaire en sortie de la couche interm\u00e9diaire choisie,\u00a0 \u00e0 celle de l&rsquo;image de contenu original.<\/b><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-acd126e elementor-widget elementor-widget-shortcode\" data-id=\"acd126e\" data-element_type=\"widget\" data-widget_type=\"shortcode.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div class=\"elementor-shortcode\"><script src=\"https:\/\/gist.github.com\/Raphkas15\/5173e6faf72d9859e53ddd4a773eec53.js\"><\/script><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-2557c2e elementor-widget elementor-widget-text-editor\" data-id=\"2557c2e\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Pour le style, en raison du nombre de couches, la fonction de perte sera diff\u00e9rente.<\/span><\/p><p><span style=\"font-weight: 400;\">Au lieu de comparer les sorties des couches interm\u00e9diaires brutes de l&rsquo;image d&rsquo;entr\u00e9e de base et de l&rsquo;image de style, les auteurs du papier original\u00a0 utilisent la diff\u00e9rence entre les matrices de Gram des couches s\u00e9lectionn\u00e9es.<\/span><\/p><p><b>La matrice de Gram est une matrice carr\u00e9e qui contient les produits des points entre chaque filtre vectoris\u00e9 de la couche. La matrice de Gram peut donc \u00eatre consid\u00e9r\u00e9e comme une matrice de corr\u00e9lation non-normalis\u00e9e pour les filtres d\u2019une couche.<\/b><\/p><p><span style=\"font-weight: 400;\">Pour g\u00e9n\u00e9rer un style pour notre image d&rsquo;entr\u00e9e, nous effectuons une <\/span><b>descente en gradient<\/b><span style=\"font-weight: 400;\"> \u00e0 partir de l&rsquo;image de contenu pour la transformer en une image qui copiera le style de l\u2019image de style. Vous ne connaissez pas la descente de gradient ? Pas de panique ! DataScientest vous l\u2019explique <a href=\"https:\/\/www.youtube.com\/watch?v=Ka4AnSXxF2g\">en vid\u00e9o<\/a> !<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-1bd6984 elementor-widget elementor-widget-shortcode\" data-id=\"1bd6984\" data-element_type=\"widget\" data-widget_type=\"shortcode.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div class=\"elementor-shortcode\"><script src=\"https:\/\/gist.github.com\/Raphkas15\/a0b38a04831b0b67d31e01fe7f443d7d.js\"><\/script><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-c1de49c elementor-widget elementor-widget-text-editor\" data-id=\"c1de49c\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Il ne nous reste plus qu\u2019\u00e0 cr\u00e9er une fonction pour charger nos diff\u00e9rentes images, les faire passer dans notre mod\u00e8le pour retourner les repr\u00e9sentations de contenu et de style, puis de calculer les fonctions de perte et le gradient afin de passer \u00e0 l\u2019optimisation<\/span><\/p><p><span style=\"font-weight: 400;\">Et voil\u00e0 !<\/span><\/p><p><span style=\"font-weight: 400;\">Nous pouvons ensuite appliquer la m\u00e9thode \u00e0 un certain nombre d\u2019images pour tester notre style transfer. En voici quelques exemples :<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-bf0891c elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"bf0891c\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-13446cf\" data-id=\"13446cf\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<section class=\"elementor-section elementor-inner-section elementor-element elementor-element-ef8aab4 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"ef8aab4\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-50 elementor-inner-column elementor-element elementor-element-6e20725\" data-id=\"6e20725\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-60f53ca elementor-widget elementor-widget-image\" data-id=\"60f53ca\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"800\" height=\"584\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/5-1-1024x747.png\" class=\"attachment-large size-large wp-image-6991\" alt=\"\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/5-1-1024x747.png 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/5-1-300x219.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/5-1-768x560.png 768w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/5-1.png 1480w\" sizes=\"(max-width: 800px) 100vw, 800px\" \/>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t<div class=\"elementor-column elementor-col-50 elementor-inner-column elementor-element elementor-element-f69ae7e\" data-id=\"f69ae7e\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-28f6fd5 elementor-widget elementor-widget-image\" data-id=\"28f6fd5\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"800\" height=\"584\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/4-1-1024x747.png\" class=\"attachment-large size-large wp-image-6990\" alt=\"\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/4-1-1024x747.png 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/4-1-300x219.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/4-1-768x560.png 768w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/4-1.png 1480w\" sizes=\"(max-width: 800px) 100vw, 800px\" \/>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-inner-section elementor-element elementor-element-e015f58 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"e015f58\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-50 elementor-inner-column elementor-element elementor-element-df26002\" data-id=\"df26002\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-5a37666 elementor-widget elementor-widget-image\" data-id=\"5a37666\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"800\" height=\"584\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/3-2-1024x747.png\" class=\"attachment-large size-large wp-image-6989\" alt=\"\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/3-2-1024x747.png 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/3-2-300x219.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/3-2-768x560.png 768w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/3-2.png 1480w\" sizes=\"(max-width: 800px) 100vw, 800px\" \/>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t<div class=\"elementor-column elementor-col-50 elementor-inner-column elementor-element elementor-element-1e2177e\" data-id=\"1e2177e\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-43ec396 elementor-widget elementor-widget-image\" data-id=\"43ec396\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"800\" height=\"584\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/2-4-1024x747.png\" class=\"attachment-large size-large wp-image-6988\" alt=\"\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/2-4-1024x747.png 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/2-4-300x219.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/2-4-768x560.png 768w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/2-4.png 1480w\" sizes=\"(max-width: 800px) 100vw, 800px\" \/>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-0b5a706 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"0b5a706\" data-element_type=\"section\">\n\t\t\t\t\t\t<div class=\"elementor-container elementor-column-gap-default\">\n\t\t\t\t\t<div class=\"elementor-column elementor-col-100 elementor-top-column elementor-element elementor-element-5496b11\" data-id=\"5496b11\" data-element_type=\"column\">\n\t\t\t<div class=\"elementor-widget-wrap elementor-element-populated\">\n\t\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-6e1bdb5 elementor-widget elementor-widget-text-editor\" data-id=\"6e1bdb5\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><strong>Vous aimeriez aller plus loin dans votre apprentissage du Deep Learning ? D\u00e9marrez prochainement l\u2019une de nos formations Data Scientist !\u00a0<\/strong><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t\t<\/div>\n\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t<\/section>\n\t\t\t\t\t\t\t<\/div>\n\t\t","protected":false},"excerpt":{"rendered":"<p>Avez-vous d\u00e9j\u00e0 souhait\u00e9 pouvoir peindre comme Van Gogh? Reproduire l\u2019esth\u00e9tique de Monet dans des paysages modernes? Dans cet article, nous allons vous pr\u00e9senter une technique utilisant le Deep Learning pour appliquer \u00e0 une image d\u2019origine le style d\u2019une autre. Cette technique d\u2019optimisation est connue sous le nom de neural style transfer (ou transfert de style), [&hellip;]<\/p>\n","protected":false},"author":78,"featured_media":26169,"comment_status":"open","ping_status":"open","sticky":false,"template":"elementor_theme","format":"standard","meta":{"content-type":"","inline_featured_image":false},"categories":[2497],"tags":[],"yoast_head":"<!-- This site is optimized with the Yoast SEO Premium plugin v21.7 (Yoast SEO v21.7) - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>Deep learning &amp; Style transfer - Formation Data Science<\/title>\n<meta name=\"description\" content=\"D\u00e9couvrez la technique d\u2019optimisation connue sous le nom de neural style transfer (ou transfert de style) gr\u00e2ce \u00e0 nos experts !\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/datascientest.com\/deep-learning-et-style-transfer\" \/>\n<meta property=\"og:locale\" content=\"fr_FR\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Deep learning &amp; Style transfer\" \/>\n<meta property=\"og:url\" content=\"https:\/\/datascientest.com\/deep-learning-et-style-transfer\" \/>\n<meta property=\"og:site_name\" content=\"Formation Data Science | DataScientest.com\" \/>\n<meta property=\"article:published_time\" content=\"2020-08-12T14:10:28+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2023-11-09T14:12:45+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/illu_DLtransfer_blog-33.png\" \/>\n\t<meta property=\"og:image:width\" content=\"2560\" \/>\n\t<meta property=\"og:image:height\" content=\"1405\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/png\" \/>\n<meta name=\"author\" content=\"J\u00e9r\u00e9my Robert\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u00c9crit par\" \/>\n\t<meta name=\"twitter:data1\" content=\"J\u00e9r\u00e9my Robert\" \/>\n\t<meta name=\"twitter:label2\" content=\"Dur\u00e9e de lecture estim\u00e9e\" \/>\n\t<meta name=\"twitter:data2\" content=\"8 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/datascientest.com\/deep-learning-et-style-transfer\",\"url\":\"https:\/\/datascientest.com\/deep-learning-et-style-transfer\",\"name\":\"Deep learning &amp; Style transfer - Formation Data Science\",\"isPartOf\":{\"@id\":\"https:\/\/datascientest.com\/#website\"},\"datePublished\":\"2020-08-12T14:10:28+00:00\",\"dateModified\":\"2023-11-09T14:12:45+00:00\",\"author\":{\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/7af2361c48e24e79a8aa2360e6e74925\"},\"description\":\"D\u00e9couvrez la technique d\u2019optimisation connue sous le nom de neural style transfer (ou transfert de style) gr\u00e2ce \u00e0 nos experts !\",\"inLanguage\":\"fr-FR\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/datascientest.com\/deep-learning-et-style-transfer\"]}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/datascientest.com\/#website\",\"url\":\"https:\/\/datascientest.com\/\",\"name\":\"Formation Data Science | DataScientest.com\",\"description\":\"Leader de la formation en Data Science\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/datascientest.com\/?s={search_term_string}\"},\"query-input\":\"required name=search_term_string\"}],\"inLanguage\":\"fr-FR\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/7af2361c48e24e79a8aa2360e6e74925\",\"name\":\"J\u00e9r\u00e9my Robert\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"fr-FR\",\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/jeremyrobert.jpeg\",\"contentUrl\":\"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/jeremyrobert.jpeg\",\"caption\":\"J\u00e9r\u00e9my Robert\"},\"description\":\"J\u00e9r\u00e9my Robert est actuellement Directeur p\u00e9dagogique B2B chez DataScientest.com. Avant de rejoindre DataScientest, il a travaill\u00e9 en tant que Consultant Business Intelligence chez DEAPDATA, o\u00f9 il a g\u00e9r\u00e9 des projets dans les domaines de l'assurance, bancaire et industrie. Il a \u00e9galement \u00e9t\u00e9 Co-fondateur et pr\u00e9sident de Tripzer. En tant que consultant ind\u00e9pendant, J\u00e9r\u00e9my a propos\u00e9 des formations sur MicroStrategy pour divers niveaux professionnels. Il est dipl\u00f4m\u00e9 de l'Universit\u00e9 Paris Cit\u00e9 avec un Master MIAGE, combinant comp\u00e9tences en informatique et gestion.\",\"sameAs\":[\"https:\/\/www.linkedin.com\/in\/jeremyrobert1\/\",\"Data Science\"],\"url\":\"https:\/\/datascientest.com\/author\/jeremyrobert\"}]}<\/script>\n<!-- \/ Yoast SEO Premium plugin. -->","yoast_head_json":{"title":"Deep learning &amp; Style transfer - Formation Data Science","description":"D\u00e9couvrez la technique d\u2019optimisation connue sous le nom de neural style transfer (ou transfert de style) gr\u00e2ce \u00e0 nos experts !","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/datascientest.com\/deep-learning-et-style-transfer","og_locale":"fr_FR","og_type":"article","og_title":"Deep learning &amp; Style transfer","og_url":"https:\/\/datascientest.com\/deep-learning-et-style-transfer","og_site_name":"Formation Data Science | DataScientest.com","article_published_time":"2020-08-12T14:10:28+00:00","article_modified_time":"2023-11-09T14:12:45+00:00","og_image":[{"width":2560,"height":1405,"url":"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/08\/illu_DLtransfer_blog-33.png","type":"image\/png"}],"author":"J\u00e9r\u00e9my Robert","twitter_card":"summary_large_image","twitter_misc":{"\u00c9crit par":"J\u00e9r\u00e9my Robert","Dur\u00e9e de lecture estim\u00e9e":"8 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/datascientest.com\/deep-learning-et-style-transfer","url":"https:\/\/datascientest.com\/deep-learning-et-style-transfer","name":"Deep learning &amp; Style transfer - Formation Data Science","isPartOf":{"@id":"https:\/\/datascientest.com\/#website"},"datePublished":"2020-08-12T14:10:28+00:00","dateModified":"2023-11-09T14:12:45+00:00","author":{"@id":"https:\/\/datascientest.com\/#\/schema\/person\/7af2361c48e24e79a8aa2360e6e74925"},"description":"D\u00e9couvrez la technique d\u2019optimisation connue sous le nom de neural style transfer (ou transfert de style) gr\u00e2ce \u00e0 nos experts !","inLanguage":"fr-FR","potentialAction":[{"@type":"ReadAction","target":["https:\/\/datascientest.com\/deep-learning-et-style-transfer"]}]},{"@type":"WebSite","@id":"https:\/\/datascientest.com\/#website","url":"https:\/\/datascientest.com\/","name":"Formation Data Science | DataScientest.com","description":"Leader de la formation en Data Science","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/datascientest.com\/?s={search_term_string}"},"query-input":"required name=search_term_string"}],"inLanguage":"fr-FR"},{"@type":"Person","@id":"https:\/\/datascientest.com\/#\/schema\/person\/7af2361c48e24e79a8aa2360e6e74925","name":"J\u00e9r\u00e9my Robert","image":{"@type":"ImageObject","inLanguage":"fr-FR","@id":"https:\/\/datascientest.com\/#\/schema\/person\/image\/","url":"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/jeremyrobert.jpeg","contentUrl":"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/jeremyrobert.jpeg","caption":"J\u00e9r\u00e9my Robert"},"description":"J\u00e9r\u00e9my Robert est actuellement Directeur p\u00e9dagogique B2B chez DataScientest.com. Avant de rejoindre DataScientest, il a travaill\u00e9 en tant que Consultant Business Intelligence chez DEAPDATA, o\u00f9 il a g\u00e9r\u00e9 des projets dans les domaines de l'assurance, bancaire et industrie. Il a \u00e9galement \u00e9t\u00e9 Co-fondateur et pr\u00e9sident de Tripzer. En tant que consultant ind\u00e9pendant, J\u00e9r\u00e9my a propos\u00e9 des formations sur MicroStrategy pour divers niveaux professionnels. Il est dipl\u00f4m\u00e9 de l'Universit\u00e9 Paris Cit\u00e9 avec un Master MIAGE, combinant comp\u00e9tences en informatique et gestion.","sameAs":["https:\/\/www.linkedin.com\/in\/jeremyrobert1\/","Data Science"],"url":"https:\/\/datascientest.com\/author\/jeremyrobert"}]}},"_links":{"self":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts\/6792"}],"collection":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/users\/78"}],"replies":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/comments?post=6792"}],"version-history":[{"count":0,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts\/6792\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/media\/26169"}],"wp:attachment":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/media?parent=6792"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/categories?post=6792"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/tags?post=6792"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}