{"id":4423,"date":"2020-06-25T15:00:33","date_gmt":"2020-06-25T14:00:33","guid":{"rendered":"https:\/\/datascientest.com\/?p=4423"},"modified":"2024-01-31T09:31:50","modified_gmt":"2024-01-31T09:31:50","slug":"convolutional-neural-network","status":"publish","type":"post","link":"https:\/\/datascientest.com\/convolutional-neural-network","title":{"rendered":"Convolutional Neural Network : Tout ce qu&rsquo;il y a \u00e0 savoir"},"content":{"rendered":"\t\t<div data-elementor-type=\"wp-post\" data-elementor-id=\"4423\" class=\"elementor elementor-4423\" data-elementor-post-type=\"post\">\n\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-9d953b4 e-flex e-con-boxed e-con\" data-id=\"9d953b4\" 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-a76eb0b elementor-widget elementor-widget-heading\" data-id=\"a76eb0b\" 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\">Dans cette partie, nous allons nous focaliser sur un des algorithmes les plus performants du Deep Learning : les&nbsp;Convolutional Neural Network&nbsp;(CNN) ou r\u00e9seaux de neurones convolutifs en fran\u00e7ais. Ce sont des mod\u00e8les de programmation puissants permettant notamment la reconnaissance d\u2019images en attribuant automatiquement \u00e0 chaque image fournie en entr\u00e9e, une \u00e9tiquette correspondant \u00e0 sa classe d\u2019appartenance.<\/h2>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-5754a08 elementor-widget elementor-widget-text-editor\" data-id=\"5754a08\" 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>Bienvenue dans le troisi\u00e8me \u00e9pisode de notre dossier Deep Learning. Apr\u00e8s avoir introduit le Deep Learning et ses applications dans la premi\u00e8re partie, nous nous sommes pench\u00e9s sur la structure et le fonctionnement des r\u00e9seaux de neurones dans la seconde.&nbsp;<span style=\"color: var( --e-global-color-text ); font-family: var( --e-global-typography-text-font-family ), Sans-serif; font-size: var( --e-global-typography-text-font-size ); font-weight: var( --e-global-typography-text-font-weight ); text-align: var(--text-align);\">En route !<\/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-47b1f96 elementor-widget__width-initial elementor-widget elementor-widget-video\" data-id=\"47b1f96\" data-element_type=\"widget\" data-settings=\"{&quot;youtube_url&quot;:&quot;https:\\\/\\\/youtu.be\\\/DazUaVu5MO0&quot;,&quot;video_type&quot;:&quot;youtube&quot;,&quot;controls&quot;:&quot;yes&quot;}\" data-widget_type=\"video.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-video .elementor-widget-container{overflow:hidden;transform:translateZ(0)}.elementor-widget-video .elementor-wrapper{aspect-ratio:var(--video-aspect-ratio)}.elementor-widget-video .elementor-wrapper iframe,.elementor-widget-video .elementor-wrapper video{height:100%;width:100%;display:flex;border:none;background-color:#000}@supports not (aspect-ratio:1\/1){.elementor-widget-video .elementor-wrapper{position:relative;overflow:hidden;height:0;padding-bottom:calc(100% \/ var(--video-aspect-ratio))}.elementor-widget-video .elementor-wrapper iframe,.elementor-widget-video .elementor-wrapper video{position:absolute;top:0;right:0;bottom:0;left:0}}.elementor-widget-video .elementor-open-inline .elementor-custom-embed-image-overlay{position:absolute;top:0;right:0;bottom:0;left:0;background-size:cover;background-position:50%}.elementor-widget-video .elementor-custom-embed-image-overlay{cursor:pointer;text-align:center}.elementor-widget-video .elementor-custom-embed-image-overlay:hover .elementor-custom-embed-play i{opacity:1}.elementor-widget-video .elementor-custom-embed-image-overlay img{display:block;width:100%;aspect-ratio:var(--video-aspect-ratio);-o-object-fit:cover;object-fit:cover;-o-object-position:center center;object-position:center center}@supports not (aspect-ratio:1\/1){.elementor-widget-video .elementor-custom-embed-image-overlay{position:relative;overflow:hidden;height:0;padding-bottom:calc(100% \/ var(--video-aspect-ratio))}.elementor-widget-video .elementor-custom-embed-image-overlay img{position:absolute;top:0;right:0;bottom:0;left:0}}.elementor-widget-video .e-hosted-video .elementor-video{-o-object-fit:cover;object-fit:cover}.e-con-inner>.elementor-widget-video,.e-con>.elementor-widget-video{width:var(--container-widget-width);--flex-grow:var(--container-widget-flex-grow)}<\/style>\t\t<div class=\"elementor-wrapper elementor-open-inline\">\n\t\t\t<div class=\"elementor-video\"><\/div>\t\t<\/div>\n\t\t\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<div class=\"elementor-element elementor-element-d0a0a1a e-flex e-con-boxed e-con\" data-id=\"d0a0a1a\" 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-04c8fc7 elementor-widget elementor-widget-text-editor\" data-id=\"04c8fc7\" 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>Table des mati\u00e8res :<\/p><ol><li><a href=\"#architecturecnn\">Architecture d\u2019un Convolutional Neural Network-CNN<\/a><\/li><li><a href=\"#applications\">Quels sont les principales applications des CNN ?<\/a><\/li><li><a href=\"#partieconvolutive\">Partie convolutive<\/a><\/li><li><a href=\"#maxpoooling\">Le Max pooling<\/a><\/li><li><a href=\"#miseenapplicationcnn\">Mise en application du CNN<\/a><\/li><\/ol>\t\t\t\t\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-d57a369 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"d57a369\" 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-9972d5e\" data-id=\"9972d5e\" 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-5b09b7d elementor-widget elementor-widget-image\" data-id=\"5b09b7d\" 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<figure class=\"wp-caption\">\n\t\t\t\t\t\t\t\t\t\t\t<a href=\"https:\/\/datascientest.com\/catalog\/module\/151-introduction-au-deeplearning-avec-le-framework-keras\">\n\t\t\t\t\t\t\t<img decoding=\"async\" width=\"370\" height=\"264\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/CNN-1.png\" class=\"attachment-full size-full wp-image-4469\" alt=\"modele cnn\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/CNN-1.png 370w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/CNN-1-300x214.png 300w\" sizes=\"(max-width: 370px) 100vw, 370px\" \/>\t\t\t\t\t\t\t\t<\/a>\n\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">Reconnaissance d\u2019un p\u00e8re-no\u00ebl par un mod\u00e8le CNN<\/figcaption>\n\t\t\t\t\t\t\t\t\t\t<\/figure>\n\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<div class=\"elementor-element elementor-element-e2f473f e-flex e-con-boxed e-con\" data-id=\"e2f473f\" 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-1a75494 elementor-widget elementor-widget-menu-anchor\" data-id=\"1a75494\" data-element_type=\"widget\" data-widget_type=\"menu-anchor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<style>\/*! elementor - v3.15.0 - 02-08-2023 *\/\nbody.elementor-page .elementor-widget-menu-anchor{margin-bottom:0}<\/style>\t\t<div id=\"applications\" class=\"elementor-menu-anchor\"><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-5906383 elementor-widget elementor-widget-heading\" data-id=\"5906383\" 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\">Quels sont les principales applications des CNN ?<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-7cdf3d8 elementor-widget elementor-widget-text-editor\" data-id=\"7cdf3d8\" 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;\">Un r\u00e9seau neuronal convolutif s\u2019inspire de la nature, car <strong>la connectivit\u00e9 entre les neurones artificiels ressemble \u00e0 l\u2019organisation du cortex visuel animal<\/strong>.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">L\u2019un des principaux cas d\u2019usage est la <strong>reconnaissance d\u2019image<\/strong>. Les r\u00e9seaux convolutifs apprennent plus rapidement et ont un meilleur taux d\u2019erreur. Dans une moindre mesure, on les utilise aussi pour l\u2019analyse vid\u00e9o.\u00a0\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">Ce type de r\u00e9seau est aussi utilis\u00e9 pour le <a href=\"https:\/\/datascientest.com\/formation-nlp\">traitement naturel du langage<\/a>. Les mod\u00e8les CNN sont tr\u00e8s efficaces pour l\u2019<strong>analyse s\u00e9mantique<\/strong>, la <strong>mod\u00e9lisation de phrase<\/strong>, la <strong>classification<\/strong> ou la <strong>traduction<\/strong>.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">Par rapport \u00e0 des m\u00e9thodes traditionnelles comme les r\u00e9seaux de neurones r\u00e9currents, les <strong>r\u00e9seaux de neurones \u00e0 convolution<\/strong> peuvent repr\u00e9senter diff\u00e9rentes r\u00e9alit\u00e9s contextuelles du langage qui ne reposent pas sur une hypoth\u00e8se de s\u00e9quence en s\u00e9rie.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">Les r\u00e9seaux convolutifs ont aussi \u00e9t\u00e9 <strong>utilis\u00e9s pour la d\u00e9couverte de m\u00e9dicaments<\/strong>. Ils permettent d\u2019identifier les traitements potentiels en pr\u00e9disant les interactions entre mol\u00e9cules et prot\u00e9ines biologiques.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">Les CNN ont notamment \u00e9t\u00e9 utilis\u00e9s <strong>pour les logiciels de jeu de Go ou de jeu d&rsquo;\u00e9checs<\/strong> auxquels ils sont capables d\u2019exceller. Une autre application est la d\u00e9tection d\u2019anomalie sur une image en entr\u00e9e (input).<\/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\t\t<\/div>\n\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-14ea57f elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"14ea57f\" 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-db37e72\" data-id=\"db37e72\" 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-e4bc4c0 elementor-widget elementor-widget-menu-anchor\" data-id=\"e4bc4c0\" data-element_type=\"widget\" data-widget_type=\"menu-anchor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div id=\"architecturecnn\" class=\"elementor-menu-anchor\"><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-ba14e68 elementor-widget elementor-widget-heading\" data-id=\"ba14e68\" 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\">Architecture d\u2019un Convolutional Neural Network-CNN <\/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-50e452d elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"50e452d\" 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-9537378\" data-id=\"9537378\" 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-9f33b3c elementor-widget elementor-widget-text-editor\" data-id=\"9f33b3c\" 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;\">Les <\/span><b>CNN<\/b><span style=\"font-weight: 400;\"> d\u00e9signent une sous-cat\u00e9gorie de r\u00e9seaux de neurones et sont \u00e0 ce jour un des <\/span><b>mod\u00e8les de classification d\u2019images<\/b><span style=\"font-weight: 400;\"> r\u00e9put\u00e9s \u00eatre les plus performant.<\/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-fba700d elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"fba700d\" 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-0ceff5f\" data-id=\"0ceff5f\" 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-579859c elementor-widget elementor-widget-text-editor\" data-id=\"579859c\" 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>\u00a0Leur mode de fonctionnement est \u00e0 premi\u00e8re vue simple : l\u2019utilisateur fournit en entr\u00e9e\u00a0<strong>une image sous la forme d\u2019une matrice de pixels<\/strong>.<br \/>Celle-ci dispose de 3 dimensions :<\/p><ul><li><strong>Deux dimensions<\/strong>\u00a0pour une image en niveaux de gris.<\/li><li>Une\u00a0<strong>troisi\u00e8me dimension<\/strong>, de profondeur 3 pour repr\u00e9senter les couleurs fondamentales (<strong>Rouge, Vert, Bleu<\/strong>).<\/li><\/ul>\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-da3c8a1 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"da3c8a1\" 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-e9a151a\" data-id=\"e9a151a\" 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-3abe7f5 elementor-widget elementor-widget-text-editor\" data-id=\"3abe7f5\" 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;\">Contrairement \u00e0 un <\/span><b>mod\u00e8le MLP (Multi Layers Perceptron)<\/b><span style=\"font-weight: 400;\"> classique qui ne contient qu\u2019une partie classification, l\u2019architecture du Convolutional Neural Network dispose en amont d\u2019une partie convolutive et comporte par cons\u00e9quent deux parties bien distinctes : <\/span><\/p><ul><li><b>Une partie convolutive <\/b><span style=\"font-weight: 400;\">: Son objectif final est d\u2019<\/span><b>extraire des caract\u00e9ristiques propres \u00e0 chaque image<\/b><span style=\"font-weight: 400;\"> en les <\/span><b>compressant <\/b><span style=\"font-weight: 400;\">de fa\u00e7on \u00e0 <\/span><b>r\u00e9duire leur taille initiale<\/b><span style=\"font-weight: 400;\">. En r\u00e9sum\u00e9, l\u2019image fournie en entr\u00e9e passe \u00e0 travers une <\/span><b>succession de filtres<\/b><span style=\"font-weight: 400;\">, cr\u00e9ant par la m\u00eame occasion de nouvelles images appel\u00e9es <\/span><b>cartes de convolutions<\/b><span style=\"font-weight: 400;\">. Enfin, les <\/span><b>cartes de convolutions obtenues sont concat\u00e9n\u00e9es<\/b><span style=\"font-weight: 400;\"> dans un vecteur de caract\u00e9ristiques appel\u00e9 <\/span><b>code CNN<\/b><span style=\"font-weight: 400;\">.<\/span><\/li><li><b>Une partie classification<\/b><span style=\"font-weight: 400;\"> : Le <\/span><b>code CNN obtenu en sortie<\/b><span style=\"font-weight: 400;\"> de la partie convolutive est <\/span><b>fourni en entr\u00e9e dans une deuxi\u00e8me partie<\/b><span style=\"font-weight: 400;\">, constitu\u00e9e de couches enti\u00e8rement connect\u00e9es appel\u00e9es perceptron multicouche (MLP pour Multi Layers Perceptron). Le r\u00f4le de cette partie est de <\/span><b>combiner les caract\u00e9ristiques du code CNN afin de classer l\u2019image<\/b><span style=\"font-weight: 400;\">. Pour revenir sur cette partie, n&rsquo;h\u00e9sitez pas \u00e0\u00a0<a href=\"https:\/\/datascientest.com\/deep-learning-reseau-de-neurones-biologiques-ou-artificiels-meme-combat\" target=\"_blank\" rel=\"noopener\">consulter l&rsquo;article sur le sujet<\/a>\u00a0.\u00a0<\/span><\/li><\/ul>\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-be1c9b1 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"be1c9b1\" 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-16ba27d\" data-id=\"16ba27d\" 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-b24faca elementor-widget elementor-widget-image\" data-id=\"b24faca\" 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<figure class=\"wp-caption\">\n\t\t\t\t\t\t\t\t\t\t\t<a href=\"https:\/\/blog.octo.com\/classification-dimages-les-reseaux-de-neurones-convolutifs-en-toute-simplicite\/\">\n\t\t\t\t\t\t\t<img decoding=\"async\" width=\"1920\" height=\"540\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/convolutif.png\" class=\"attachment-full size-full wp-image-188603\" alt=\"\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/convolutif.png 1920w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/convolutif-300x84.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/convolutif-1024x288.png 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/convolutif-768x216.png 768w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/convolutif-1536x432.png 1536w\" sizes=\"(max-width: 1920px) 100vw, 1920px\" \/>\t\t\t\t\t\t\t\t<\/a>\n\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">Sch\u00e9ma repr\u00e9sentant l\u2019architecture d\u2019un CNN<\/figcaption>\n\t\t\t\t\t\t\t\t\t\t<\/figure>\n\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-7dbfda2 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"7dbfda2\" 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-a4ec15e\" data-id=\"a4ec15e\" 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-ca5e0f5 elementor-widget elementor-widget-menu-anchor\" data-id=\"ca5e0f5\" data-element_type=\"widget\" data-widget_type=\"menu-anchor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div id=\"partieconvolutive\" class=\"elementor-menu-anchor\"><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-f421e61 elementor-widget elementor-widget-heading\" data-id=\"f421e61\" 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\">Partie convolutive : \u00c0 quoi sert la convolution ?<\/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-9cc2afe elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"9cc2afe\" 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-2756347\" data-id=\"2756347\" 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-c45282b elementor-widget elementor-widget-heading\" data-id=\"c45282b\" 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\">Tout d\u2019abord, \u00e0 quoi sert la convolution ?<\/h4>\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-c91f422 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"c91f422\" 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-45915f5\" data-id=\"45915f5\" 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-42318c8 elementor-widget elementor-widget-text-editor\" data-id=\"42318c8\" 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<\/span><b> convolution <\/b><span style=\"font-weight: 400;\">est une op\u00e9ration math\u00e9matique simple g\u00e9n\u00e9ralement utilis\u00e9e pour le <\/span><b>traitement et la reconnaissance d\u2019images.<\/b><span style=\"font-weight: 400;\"> Sur une image, son effet s\u2019assimile \u00e0 un filtrage dont voici le fonctionnement\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-c4519e2 elementor-widget elementor-widget-image\" data-id=\"c4519e2\" 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<figure class=\"wp-caption\">\n\t\t\t\t\t\t\t\t\t\t\t<a href=\"https:\/\/www.natural-solutions.eu\/blog\/la-reconnaissance-dimage-avec-les-rseaux-de-neurones-convolutifs#:~:text=Qu&#039;est%2Dce%20que%20la,%C3%A0%20la%20reconnaissance%20d&#039;image.\">\n\t\t\t\t\t\t\t<img decoding=\"async\" width=\"1201\" height=\"744\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Fichier-98.png\" class=\"attachment-full size-full wp-image-188604\" alt=\"\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Fichier-98.png 1201w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Fichier-98-300x186.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Fichier-98-1024x634.png 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Fichier-98-768x476.png 768w\" sizes=\"(max-width: 1201px) 100vw, 1201px\" \/>\t\t\t\t\t\t\t\t<\/a>\n\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">Sch\u00e9ma du parcours de la fen\u00eatre de filtre sur l'image<\/figcaption>\n\t\t\t\t\t\t\t\t\t\t<\/figure>\n\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-1774dd2 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"1774dd2\" 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-c340912\" data-id=\"c340912\" 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-3034022 elementor-widget elementor-widget-text-editor\" data-id=\"3034022\" 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<ol><li><span style=\"font-weight: 400;\">Dans un premier temps, on <\/span><b>d\u00e9finit la taille de la fen\u00eatre de filtre <\/b><span style=\"font-weight: 400;\">situ\u00e9e en haut \u00e0 gauche. <\/span><\/li><li><span style=\"font-weight: 400;\">La fen\u00eatre de filtre, repr\u00e9sentant la feature, <\/span><b>se d\u00e9place progressivement<\/b><span style=\"font-weight: 400;\"> de la <\/span><b>gauche vers la droite<\/b><span style=\"font-weight: 400;\"> d\u2019un certain nombre de cases d\u00e9fini au pr\u00e9alable (<\/span><b>le pas<\/b><span style=\"font-weight: 400;\">) jusqu\u2019\u00e0 arriver au bout de l\u2019image.<\/span><\/li><li><span style=\"font-weight: 400;\">\u00c0 chaque portion d\u2019image rencontr\u00e9e, un calcul de convolution s\u2019effectue permettant d\u2019obtenir en sortie une <\/span><b>carte d\u2019activation <\/b><span style=\"font-weight: 400;\">ou <\/span><b>feature map<\/b><span style=\"font-weight: 400;\"> qui indique o\u00f9 est localis\u00e9es les features dans l\u2019image : plus la feature map est \u00e9lev\u00e9e, plus la portion de l\u2019image balay\u00e9e ressemble \u00e0 la feature.<\/span><\/li><\/ol>\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-4824bae elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"4824bae\" 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-445ac93\" data-id=\"445ac93\" 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-e382902 elementor-widget elementor-widget-heading\" data-id=\"e382902\" 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\">Exemple d\u2019un filtre de convolution classique \n<\/h4>\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-2c03b12 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"2c03b12\" 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-93b60e7\" data-id=\"93b60e7\" 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-8d38ee8 elementor-widget elementor-widget-text-editor\" data-id=\"8d38ee8\" 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;\">Lors de la partie convolutive d\u2019un Convolutional Neural Network, l\u2019image fournie en entr\u00e9e passe \u00e0 travers une <\/span><b>succession de filtres de convolution<\/b><span style=\"font-weight: 400;\">. Par exemple,<\/span> <span style=\"font-weight: 400;\">il existe des filtres de convolution fr\u00e9quemment utilis\u00e9s et permettant\u00a0d\u2019extraire des caract\u00e9ristiques plus pertinentes que des pixels comme la d\u00e9tection des bords (<\/span><b>filtre d\u00e9rivateur<\/b><span style=\"font-weight: 400;\">) ou des formes g\u00e9om\u00e9triques. Le choix et l\u2019application des filtres se fait <\/span><b>automatiquement<\/b><span style=\"font-weight: 400;\"> par le mod\u00e8le.<\/span><\/p><p><span style=\"font-weight: 400;\">Parmi les filtres les plus connus, on retrouve notamment le <\/span><b>filtre moyenneur<\/b><span style=\"font-weight: 400;\"> (calcule pour chaque pixel la moyenne du pixel avec ses 8 proches voisins) ou encore le\u00a0<\/span><b>filtre gaussien<\/b><span style=\"font-weight: 400;\"> permettant de r\u00e9duire le bruit d\u2019une image fournie en entr\u00e9e :<\/span><\/p><p><span style=\"font-weight: 400;\">Voici un exemple des effets de ces deux diff\u00e9rents filtres sur une image comportant un bruit important (on peut penser \u00e0 une photographie prise avec une faible luminosit\u00e9 par exemple). Toutefois, un des inconv\u00e9nients de la r\u00e9duction du bruit est qu\u2019elle s&rsquo;accompagne g\u00e9n\u00e9ralement d&rsquo;une r\u00e9duction de la nettet\u00e9\u00a0:<\/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-8a8e100 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"8a8e100\" 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-6541140\" data-id=\"6541140\" 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-36152f0 elementor-widget elementor-widget-image\" data-id=\"36152f0\" 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<figure class=\"wp-caption\">\n\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"512\" height=\"306\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/CNN-4.png\" class=\"attachment-full size-full wp-image-4466\" alt=\"CNN Deep learning\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/CNN-4.png 512w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/CNN-4-300x179.png 300w\" sizes=\"(max-width: 512px) 100vw, 512px\" \/>\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">Effet des filtres moyenneur et gaussien - DataScientest<\/figcaption>\n\t\t\t\t\t\t\t\t\t\t<\/figure>\n\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-35db3e3 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"35db3e3\" 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-202dec2\" data-id=\"202dec2\" 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-2194949 elementor-widget elementor-widget-text-editor\" data-id=\"2194949\" 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;\">Comme on peut l\u2019observer,<\/span> <span style=\"font-weight: 400;\">contrairement au filtre moyenneur, le filtre gaussien r\u00e9duit le bruit sans pour autant r\u00e9duire significativement la nettet\u00e9.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">Outre sa fonction de filtrage, l\u2019int\u00e9r\u00eat de la <\/span><b>partie convolutive d\u2019un CNN <\/b><span style=\"font-weight: 400;\">est qu\u2019elle permet d\u2019<\/span><b>extraire des caract\u00e9ristiques propres \u00e0 chaque image <\/b><span style=\"font-weight: 400;\">en les compressant de fa\u00e7on \u00e0 r\u00e9duire leur taille initiale, via des m\u00e9thodes de sous-\u00e9chantillonnage tel que le Max-Pooling.<\/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-9e7fd1d elementor-align-center button-principale elementor-widget elementor-widget-button\" data-id=\"9e7fd1d\" data-element_type=\"widget\" data-widget_type=\"button.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div class=\"elementor-button-wrapper\">\n\t\t\t<a class=\"elementor-button elementor-button-link elementor-size-sm elementor-animation-grow\" href=\"https:\/\/datascientest.com\/formation-data-scientist\">\n\t\t\t\t\t\t<span class=\"elementor-button-content-wrapper\">\n\t\t\t\t\t\t<span class=\"elementor-button-text\">Devenir Data Scientist<\/span>\n\t\t<\/span>\n\t\t\t\t\t<\/a>\n\t\t<\/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-4b1b32e elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"4b1b32e\" 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-7bb509d\" data-id=\"7bb509d\" 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-1568211 elementor-widget elementor-widget-menu-anchor\" data-id=\"1568211\" data-element_type=\"widget\" data-widget_type=\"menu-anchor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div id=\"maxpoooling\" class=\"elementor-menu-anchor\"><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-8409d84 elementor-widget elementor-widget-heading\" data-id=\"8409d84\" 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\">M\u00e9thode de sous \u00e9chantillonnage : le Max-Pooling<\/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-410250b elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"410250b\" 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-519070a\" data-id=\"519070a\" 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-dc35588 elementor-widget elementor-widget-text-editor\" data-id=\"dc35588\" 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;\">Le Max-Pooling est un processus de discr\u00e9tisation bas\u00e9 sur des \u00e9chantillons. Son objectif est de sous-\u00e9chantillonner une repr\u00e9sentation d&rsquo;entr\u00e9e (image, matrice de sortie de couche cach\u00e9e, etc.) en r\u00e9duisant sa dimension. De plus, son int\u00e9r\u00eat est qu\u2019il r\u00e9duit le co\u00fbt de calcul en <\/span><b>r\u00e9duisant le nombre de param\u00e8tres \u00e0 apprendre<\/b><span style=\"font-weight: 400;\"> et fournit une <\/span><b>invariance par petites translations<\/b><span style=\"font-weight: 400;\"> (si une petite translation ne modifie pas le maximum de la r\u00e9gion balay\u00e9e, le maximum de chaque r\u00e9gion restera le m\u00eame et donc la nouvelle matrice cr\u00e9\u00e9e restera identique).<\/span><\/p><p><span style=\"font-weight: 400;\">Pour rendre plus concret l\u2019action du Max-Pooling, voici un exemple : imaginons que nous avons une matrice 4&#215;4 repr\u00e9sentant notre entr\u00e9e initiale et un filtre d\u2019une fen\u00eatre de taille 2&#215;2 que nous appliquerons sur notre entr\u00e9e. Pour chacune des r\u00e9gions balay\u00e9es par le filtre, le max-pooling prendra le maximum, cr\u00e9ant ainsi par la m\u00eame occasion une nouvelle matrice de sortie o\u00f9 chaque \u00e9l\u00e9ment correspondra aux maximums de chaque r\u00e9gion rencontr\u00e9e.<\/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-3c537ab elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"3c537ab\" 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-7996df4\" data-id=\"7996df4\" 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-0875309 elementor-widget elementor-widget-text-editor\" data-id=\"0875309\" 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;\">Illustrons le processus :<\/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-cb0c8e0 elementor-widget elementor-widget-image\" data-id=\"cb0c8e0\" 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<figure class=\"wp-caption\">\n\t\t\t\t\t\t\t\t\t\t\t<a href=\"https:\/\/cs231n.github.io\/convolutional-networks\/\">\n\t\t\t\t\t\t\t<img decoding=\"async\" width=\"1201\" height=\"543\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Fichier-97.png\" class=\"attachment-full size-full wp-image-188605\" alt=\"\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Fichier-97.png 1201w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Fichier-97-300x136.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Fichier-97-1024x463.png 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Fichier-97-768x347.png 768w\" sizes=\"(max-width: 1201px) 100vw, 1201px\" \/>\t\t\t\t\t\t\t\t<\/a>\n\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">Processus de Max-Pooling<\/figcaption>\n\t\t\t\t\t\t\t\t\t\t<\/figure>\n\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-3cedbf8 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"3cedbf8\" 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-ec9661a\" data-id=\"ec9661a\" 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-c225c39 elementor-widget elementor-widget-text-editor\" data-id=\"c225c39\" 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 fen\u00eatre de filtre se d\u00e9place de deux pixels vers la droite (stride\/pas = 2) et r\u00e9cup\u00e8re \u00e0 chaque pas \u201cl\u2019argmax\u201d correspondant \u00e0 la valeur la plus grande parmi les 4 valeurs de pixels.<\/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-fa3e76f elementor-widget elementor-widget-image\" data-id=\"fa3e76f\" 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<figure class=\"wp-caption\">\n\t\t\t\t\t\t\t\t\t\t\t<a href=\"https:\/\/cs231n.github.io\/convolutional-networks\/\">\n\t\t\t\t\t\t\t<img decoding=\"async\" width=\"806\" height=\"598\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Fichier-96.png\" class=\"attachment-full size-full wp-image-188607\" alt=\"\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Fichier-96.png 806w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Fichier-96-300x223.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Fichier-96-768x570.png 768w\" sizes=\"(max-width: 806px) 100vw, 806px\" \/>\t\t\t\t\t\t\t\t<\/a>\n\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">Exemple d\u2019effet du Max-Pooling<\/figcaption>\n\t\t\t\t\t\t\t\t\t\t<\/figure>\n\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-3e2f893 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"3e2f893\" 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-d4839e5\" data-id=\"d4839e5\" 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-74005e7 elementor-widget elementor-widget-text-editor\" data-id=\"74005e7\" 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>On comprend mieux l\u2019utilit\u00e9 de la partie convolutive d\u2019un CNN<\/b><span style=\"font-weight: 400;\">&nbsp;: contrairement \u00e0 un mod\u00e8le MLP classique, l\u2019ajout en amont de la partie convolutive permet d\u2019<\/span><b>obtenir en sortie une \u00ab\u00a0carte de caract\u00e9ristiques\u00a0\u00bb ou \u00ab\u00a0code CNN\u00a0\u00bb<\/b><span style=\"font-weight: 400;\"> (<\/span><i><span style=\"font-weight: 400;\">matrice de pixels situ\u00e9e \u00e0 droite dans l\u2019exemple<\/span><\/i><span style=\"font-weight: 400;\">) dont les <\/span><b>dimensions<\/b><span style=\"font-weight: 400;\"> sont <\/span><b>plus petites <\/b><span style=\"font-weight: 400;\">que celles de l\u2019image initiale ce qui va avoir l\u2019avantage de <\/span><b>diminuer <\/b><span style=\"font-weight: 400;\">grandement le nombre de <\/span><b>param\u00e8tres \u00e0 calculer dans le mod\u00e8le.<\/b><\/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-28729d8 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"28729d8\" 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-07d11d8\" data-id=\"07d11d8\" 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-8e00637 elementor-widget elementor-widget-heading\" data-id=\"8e00637\" 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\">Exemple d\u2019une architecture CNN et sa sortie<\/h4>\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-8693802 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"8693802\" 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-51d2faf\" data-id=\"51d2faf\" 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-ca003c0 elementor-widget elementor-widget-text-editor\" data-id=\"ca003c0\" 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;\">Apr\u00e8s la partie convolutive d\u2019un CNN, vient la partie classification. Cette partie classification, commun \u00e0 tous les mod\u00e8les de r\u00e9seaux de neurones, correspondant \u00e0 un\u00a0<\/span><a href=\"https:\/\/datascientest.com\/deep-learning-reseau-de-neurones-biologiques-ou-artificiels-meme-combat\" target=\"_blank\" rel=\"noopener\">mod\u00e8le perceptron multicouche (MLP<\/a><span style=\"font-weight: 400;\">).\u00a0 <\/span><\/p><p><span style=\"font-weight: 400;\">Son objectif est\u00a0 d\u2019attribuer \u00e0 chaque \u00e9chantillon de donn\u00e9es une \u00e9tiquette d\u00e9crivant sa classe d\u2019appartenance.\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-2bb77e9 elementor-widget elementor-widget-image\" data-id=\"2bb77e9\" 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<figure class=\"wp-caption\">\n\t\t\t\t\t\t\t\t\t\t\t<a href=\"https:\/\/www.natural-solutions.eu\/blog\/histoire-du-deep-learning#:~:text=Les%20perceptrons%20multicouches%3A%20MLP%20(1986)&#038;text=Les%20perceptrons%20multicouches%20ont%20pour,mieux%20cette%20base%20de%20donn%C3%A9es.\">\n\t\t\t\t\t\t\t<img decoding=\"async\" width=\"1033\" height=\"767\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Fichier-95.png\" class=\"attachment-full size-full wp-image-188608\" alt=\"\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Fichier-95.png 1033w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Fichier-95-300x223.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Fichier-95-1024x760.png 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Fichier-95-768x570.png 768w\" sizes=\"(max-width: 1033px) 100vw, 1033px\" \/>\t\t\t\t\t\t\t\t<\/a>\n\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">Repr\u00e9sentation d\u2019un perceptron multicouche<\/figcaption>\n\t\t\t\t\t\t\t\t\t\t<\/figure>\n\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-fff7dc6 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"fff7dc6\" 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-80bcd15\" data-id=\"80bcd15\" 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-e485c6b elementor-widget elementor-widget-text-editor\" data-id=\"e485c6b\" 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;\">L\u2019algorithme que les perceptrons utilisent pour mettre \u00e0 jour leurs poids (ou coefficients de r\u00e9seaux) s\u2019appelle <\/span><b>la r\u00e9tropropagation du gradient de l\u2019erreur, <\/b><span style=\"font-weight: 400;\">c\u00e9l\u00e8bre algorithme de descente de gradient que nous verrons plus en d\u00e9tail par la suite .\u00a0\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-506d91e elementor-widget elementor-widget-image\" data-id=\"506d91e\" 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<figure class=\"wp-caption\">\n\t\t\t\t\t\t\t\t\t\t\t<a href=\"http:\/\/dspace.univ-tlemcen.dz\/bitstream\/112\/12235\/1\/Classification-des-images-avec-les-reseaux-de-neurones.pdf\">\n\t\t\t\t\t\t\t<img decoding=\"async\" width=\"2033\" height=\"540\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/convolution-pooling.png\" class=\"attachment-full size-full wp-image-188609\" alt=\"\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/convolution-pooling.png 2033w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/convolution-pooling-300x80.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/convolution-pooling-1024x272.png 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/convolution-pooling-768x204.png 768w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/convolution-pooling-1536x408.png 1536w\" sizes=\"(max-width: 2033px) 100vw, 2033px\" \/>\t\t\t\t\t\t\t\t<\/a>\n\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">Exemple d\u2019architecture d\u2019un CNN<\/figcaption>\n\t\t\t\t\t\t\t\t\t\t<\/figure>\n\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-c9a9c28 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"c9a9c28\" 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-75ab1fe\" data-id=\"75ab1fe\" 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-5579c36 elementor-widget elementor-widget-text-editor\" data-id=\"5579c36\" 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;\">G\u00e9n\u00e9ralement, l\u2019<\/span><b>architecture d\u2019un Convolutional Neural Network<\/b><span style=\"font-weight: 400;\"> est sensiblement la m\u00eame :\u00a0<\/span><\/p><ul><li style=\"font-weight: 400;\"><b>Couche de convolution (CONV) : <\/b><span style=\"font-weight: 400;\">Le r\u00f4le de cette premi\u00e8re couche est d\u2019<\/span><b>analyser les images fournies en entr\u00e9e<\/b><span style=\"font-weight: 400;\"> et de <\/span><b>d\u00e9tecter la pr\u00e9sence d\u2019un ensemble de features<\/b><span style=\"font-weight: 400;\">. On obtient en sortie de cette couche un ensemble de features maps (<\/span><i><span style=\"font-weight: 400;\">cf plus haut\u00a0: \u00e0 quoi sert la convolution\u00a0?)<\/span><\/i><span style=\"font-weight: 400;\">.\u00a0<\/span><\/li><li style=\"font-weight: 400;\"><b>Couche de Pooling (POOL) :<\/b><span style=\"font-weight: 400;\"> La couche de Pooling est une op\u00e9ration g\u00e9n\u00e9ralement appliqu\u00e9e entre deux couches de convolution. Celle-ci re\u00e7oit en entr\u00e9e les features maps form\u00e9es en sortie de la couche de convolution et son r\u00f4le est de <\/span><b>r\u00e9duire la taille des images, tout en pr\u00e9servant leurs caract\u00e9ristiques les plus essentielles<\/b><span style=\"font-weight: 400;\">. Parmi les plus utilis\u00e9s, on retrouve le <\/span><b>max-pooling<\/b><span style=\"font-weight: 400;\"> mentionn\u00e9 pr\u00e9c\u00e9demment ou encore l&rsquo;<\/span><b>average pooling <\/b><span style=\"font-weight: 400;\">dont l\u2019op\u00e9ration consiste \u00e0 conserver \u00e0 chaque pas, la valeur moyenne de la fen\u00eatre de filtre.\u00a0<\/span><\/li><\/ul><blockquote><p><span style=\"font-weight: 400;\">Finalement, on obtient en sortie de cette couche de Pooling, le m\u00eame nombre de feature maps qu&rsquo;en entr\u00e9e mais consid\u00e9rablement compress\u00e9es.<\/span><\/p><\/blockquote><ul><li style=\"font-weight: 400;\"><b>La couche d\u2019activation ReLU (Rectified Linear Units) :<\/b><span style=\"font-weight: 400;\"> Cette couche<\/span><b> remplace toutes les valeurs n\u00e9gatives re\u00e7ues en entr\u00e9es par des z\u00e9ros<\/b><span style=\"font-weight: 400;\">. L\u2019int\u00e9r\u00eat de ces couches d\u2019activation est de rendre le mod\u00e8le non lin\u00e9aire et de ce fait plus complexe.<\/span><\/li><\/ul>\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-e904e28 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"e904e28\" 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-7de78de\" data-id=\"7de78de\" 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-14a63ba elementor-widget elementor-widget-image\" data-id=\"14a63ba\" 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<figure class=\"wp-caption\">\n\t\t\t\t\t\t\t\t\t\t\t<a href=\"https:\/\/ml4a.github.io\/ml4a\/fr\/neural_networks\/\">\n\t\t\t\t\t\t\t<img decoding=\"async\" width=\"274\" height=\"228\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/CNN-9.png\" class=\"attachment-full size-full wp-image-4471\" alt=\"ReLU Rectified Linear Units\" \/>\t\t\t\t\t\t\t\t<\/a>\n\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">Fonction d\u2019activation ReLU<\/figcaption>\n\t\t\t\t\t\t\t\t\t\t<\/figure>\n\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-2ba61ec elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"2ba61ec\" 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-ff37bb2\" data-id=\"ff37bb2\" 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-faff565 elementor-widget elementor-widget-text-editor\" data-id=\"faff565\" 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<ul><li><b>Couche Fully Connected (FC) :<\/b><span style=\"font-weight: 400;\">\u00a0 <\/span><span style=\"font-weight: 400;\">Ces couches sont plac\u00e9es en fin d&rsquo;architecture de CNN et sont <\/span><b>enti\u00e8rement connect\u00e9es \u00e0 tous les neurones de sorties <\/b><span style=\"font-weight: 400;\">(d\u2019o\u00f9 le terme fully-connected). <\/span><span style=\"font-weight: 400;\">Apr\u00e8s avoir re\u00e7u un vecteur en entr\u00e9e, la couche FC applique successivement une <\/span><b>combinaison lin\u00e9aire <\/b><span style=\"font-weight: 400;\">puis une <\/span><b>fonction d&rsquo;activation<\/b><span style=\"font-weight: 400;\"> dans le but<\/span> <span style=\"font-weight: 400;\">final de <\/span><b>classifier l&rsquo;input image <\/b><span style=\"font-weight: 400;\">(<\/span><i><span style=\"font-weight: 400;\">voir sch\u00e9ma suivant<\/span><\/i><span style=\"font-weight: 400;\">). <\/span><span style=\"font-weight: 400;\">Elle renvoie enfin en sortie un <\/span><b>vecteur de taille d correspondant au nombre de classes<\/b><span style=\"font-weight: 400;\"> dans lequel chaque composante repr\u00e9sente la probabilit\u00e9 pour l\u2019input image d&rsquo;appartenir \u00e0 une classe.<\/span><\/li><\/ul>\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-76b0287 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"76b0287\" 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-22b647b\" data-id=\"22b647b\" 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-9413892 elementor-widget elementor-widget-image\" data-id=\"9413892\" 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<figure class=\"wp-caption\">\n\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"512\" height=\"295\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/CNN-10.png\" class=\"attachment-full size-full wp-image-4470\" alt=\"couche fully connected\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/CNN-10.png 512w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/CNN-10-300x173.png 300w\" sizes=\"(max-width: 512px) 100vw, 512px\" \/>\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">Fonctionnement d\u2019un r\u00e9seau neuronal \u00e0 2 couches cach\u00e9es<\/figcaption>\n\t\t\t\t\t\t\t\t\t\t<\/figure>\n\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-23dc1ce elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"23dc1ce\" 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-ccc656f\" data-id=\"ccc656f\" 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-c0f205a elementor-widget elementor-widget-heading\" data-id=\"c0f205a\" 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\">Impl\u00e9mentation d\u2019un CNN pr\u00e9-entra\u00een\u00e9 sur Python : <\/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-695284e elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"695284e\" 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-e57fc50\" data-id=\"e57fc50\" 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-3c6a601 elementor-widget elementor-widget-text-editor\" data-id=\"3c6a601\" 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 des usages pratiques et \u00e9tant donn\u00e9 la complexit\u00e9 de cr\u00e9ation de <\/span><b>performants <\/b><span style=\"font-weight: 400;\">CNN \u201cfait \u00e0 la main\u201c, nous allons utiliser des r\u00e9seaux pr\u00e9-entra\u00een\u00e9s disponibles dans le module <a href=\"https:\/\/pytorch.org\/vision\/stable\/index.html\">Torchvision<\/a><\/span><span style=\"font-weight: 400;\">. Voyons comment celui-ci peut \u00eatre impl\u00e9ment\u00e9 sur Python :<\/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-ecedcff elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"ecedcff\" 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-490a3f0\" data-id=\"490a3f0\" 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-199db5e elementor-widget elementor-widget-menu-anchor\" data-id=\"199db5e\" data-element_type=\"widget\" data-widget_type=\"menu-anchor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div id=\"miseenapplicationcnn\" class=\"elementor-menu-anchor\"><\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-c1aef0a elementor-widget elementor-widget-heading\" data-id=\"c1aef0a\" 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\">Mise en application d\u2019un CNN : Identification d\u2019images quelconques \u00e0 partir du dataset ImageNet : dix millions d\u2019images \u00e9tiquet\u00e9es<\/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-6e54b6b elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"6e54b6b\" 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-cb7cab8\" data-id=\"cb7cab8\" 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-d13e04d elementor-widget elementor-widget-text-editor\" data-id=\"d13e04d\" 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;\">ImageNet est une base de donn\u00e9es de plus de dix millions d&rsquo;images \u00e9tiquet\u00e9es produit par l&rsquo;organisation du m\u00eame nom, \u00e0 destination des travaux de recherche en vision par ordinateur.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">Voici un extrait de ce volumineux dataset\u00a0:<\/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-cfbe4ca elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"cfbe4ca\" 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-edab4ea\" data-id=\"edab4ea\" 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-4ea9e21 elementor-widget elementor-widget-image\" data-id=\"4ea9e21\" 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<figure class=\"wp-caption\">\n\t\t\t\t\t\t\t\t\t\t\t<a href=\"https:\/\/machinelearningmastery.com\/use-pre-trained-vgg-model-classify-objects-photographs\/\">\n\t\t\t\t\t\t\t<img decoding=\"async\" width=\"442\" height=\"272\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/CNN-11.png\" class=\"attachment-full size-full wp-image-4479\" alt=\"Extrait de la base de donn\u00e9es ImageNet\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/CNN-11.png 442w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/CNN-11-300x185.png 300w\" sizes=\"(max-width: 442px) 100vw, 442px\" \/>\t\t\t\t\t\t\t\t<\/a>\n\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">Extrait de la base de donn\u00e9es ImageNet<\/figcaption>\n\t\t\t\t\t\t\t\t\t\t<\/figure>\n\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-3b20fb6 elementor-align-center button-principale elementor-widget elementor-widget-button\" data-id=\"3b20fb6\" data-element_type=\"widget\" data-widget_type=\"button.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t<div class=\"elementor-button-wrapper\">\n\t\t\t<a class=\"elementor-button elementor-button-link elementor-size-sm elementor-animation-grow\" href=\"https:\/\/datascientest.com\/formation-data-scientist\">\n\t\t\t\t\t\t<span class=\"elementor-button-content-wrapper\">\n\t\t\t\t\t\t<span class=\"elementor-button-text\">D\u00e9couvrir notre formation de Data Scientist<\/span>\n\t\t<\/span>\n\t\t\t\t\t<\/a>\n\t\t<\/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-041a00a elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"041a00a\" 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-be0e20f\" data-id=\"be0e20f\" 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-aab94b7 elementor-widget elementor-widget-text-editor\" data-id=\"aab94b7\" 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>\u00c9tape 0 : <\/b><span style=\"font-weight: 400;\">Tout d&rsquo;abord, importons toutes les librairies qui nous serons n\u00e9cessaires pour la suite<\/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-0807104 elementor-widget elementor-widget-shortcode\" data-id=\"0807104\" 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\/garybedoucha\/fa94546ddc37a047de6c3135b9d726ef.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-986d1d3 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"986d1d3\" 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-25b7998\" data-id=\"25b7998\" 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-60ff47b elementor-widget elementor-widget-text-editor\" data-id=\"60ff47b\" 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>\u00c9tape 1 :<\/b><span style=\"font-weight: 400;\"> Entrainement du mod\u00e8le pr\u00e9-entrain\u00e9 VGG16<\/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-5fa27c3 elementor-widget elementor-widget-shortcode\" data-id=\"5fa27c3\" 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\/garybedoucha\/4b4bf05a7373eb02de7279fd09c1c065.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-0cf6deb elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"0cf6deb\" 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-7f3dde5\" data-id=\"7f3dde5\" 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-a5150ba elementor-widget elementor-widget-text-editor\" data-id=\"a5150ba\" 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>\u00c9tape 2 :<\/b><span style=\"font-weight: 400;\"> Importation des 3 images \u00e0 pr\u00e9dire<\/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-086233a elementor-widget elementor-widget-shortcode\" data-id=\"086233a\" 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\/garybedoucha\/cef25c1c0944feefdff693be31c9d70e.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-6a05dc3 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"6a05dc3\" 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-33 elementor-top-column elementor-element elementor-element-dd58b01\" data-id=\"dd58b01\" 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-a3fdef0 elementor-widget elementor-widget-image\" data-id=\"a3fdef0\" 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.amazon.fr\/Ballon-football-Taille-standard-blanc\/dp\/B018URJZPK\">\n\t\t\t\t\t\t\t<img decoding=\"async\" width=\"275\" height=\"183\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/CNN-12.png\" class=\"attachment-large size-large wp-image-4478\" alt=\"\" \/>\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<div class=\"elementor-column elementor-col-33 elementor-top-column elementor-element elementor-element-97e464d\" data-id=\"97e464d\" 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-6c90ec0 elementor-widget elementor-widget-image\" data-id=\"6c90ec0\" 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:\/\/fr.wikipedia.org\/wiki\/Canis_lupus\">\n\t\t\t\t\t\t\t<img decoding=\"async\" width=\"800\" height=\"533\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Canis_lupus_Ernstbrunn-1024x682.jpg\" class=\"attachment-large size-large wp-image-191252\" alt=\"\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Canis_lupus_Ernstbrunn-1024x682.jpg 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Canis_lupus_Ernstbrunn-300x200.jpg 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Canis_lupus_Ernstbrunn-768x512.jpg 768w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Canis_lupus_Ernstbrunn.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\t<\/div>\n\t\t<\/div>\n\t\t\t\t<div class=\"elementor-column elementor-col-33 elementor-top-column elementor-element elementor-element-00c823e\" data-id=\"00c823e\" 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-3ba997e elementor-widget elementor-widget-image\" data-id=\"3ba997e\" 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:\/\/fr.wikipedia.org\/wiki\/General_Atomics_MQ-9_Reaper\">\n\t\t\t\t\t\t\t<img decoding=\"async\" width=\"150\" height=\"150\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/CNN-14-150x150.png\" class=\"attachment-thumbnail size-thumbnail wp-image-4476\" alt=\"\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/CNN-14-150x150.png 150w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/CNN-14-300x300.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/CNN-14.png 329w\" sizes=\"(max-width: 150px) 100vw, 150px\" \/>\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-8953aec elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"8953aec\" 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-292a98e\" data-id=\"292a98e\" 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-6c50679 elementor-widget elementor-widget-text-editor\" data-id=\"6c50679\" 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>\u00c9tape 3 :<\/b><span style=\"font-weight: 400;\"> Pr\u00e9traitement 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-1d6c78b elementor-widget elementor-widget-shortcode\" data-id=\"1d6c78b\" 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\/garybedoucha\/2d4c94833f521261a9aec7bdcf64d186.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-f5e27b4 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"f5e27b4\" 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-44d2aae\" data-id=\"44d2aae\" 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-485a3e6 elementor-widget elementor-widget-text-editor\" data-id=\"485a3e6\" 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>\u00c9tape 4 :<\/b><span style=\"font-weight: 400;\"> Pr\u00e9diction du mod\u00e8le<\/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-0071d3a elementor-widget elementor-widget-shortcode\" data-id=\"0071d3a\" 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\/garybedoucha\/1cf7fdd70574e0f61ff286dc00ed9c50.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-c620bde elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"c620bde\" 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-5faa4b3\" data-id=\"5faa4b3\" 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-2b57f28 elementor-widget elementor-widget-image\" data-id=\"2b57f28\" 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<figure class=\"wp-caption\">\n\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"592\" height=\"60\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/cnn.png\" class=\"attachment-full size-full wp-image-4480\" alt=\"prediction algorithme\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/cnn.png 592w, https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/cnn-300x30.png 300w\" sizes=\"(max-width: 592px) 100vw, 592px\" \/>\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">Affichage des pr\u00e9dictions du mod\u00e8le VGG16<\/figcaption>\n\t\t\t\t\t\t\t\t\t\t<\/figure>\n\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-d88e116 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"d88e116\" 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-adf0df1\" data-id=\"adf0df1\" 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-4a61045 elementor-widget elementor-widget-text-editor\" data-id=\"4a61045\" 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;\">On peut \u00e9galement afficher les 5 labels jug\u00e9s les plus probables par VGG16\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-5e683f3 elementor-widget elementor-widget-shortcode\" data-id=\"5e683f3\" 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\/garybedoucha\/5c0c858cc543f0f3f4a9babc4149fa93.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-1715fd1 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"1715fd1\" 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-9c94eb2\" data-id=\"9c94eb2\" 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-5e621c7 elementor-widget elementor-widget-image\" data-id=\"5e621c7\" 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<figure class=\"wp-caption\">\n\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"226\" height=\"168\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/CNN-15-.png\" class=\"attachment-full size-full wp-image-4475\" alt=\"prediction algorithme\" \/>\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\"> Top 5 des labels jug\u00e9s les plus probables par VGG16<\/figcaption>\n\t\t\t\t\t\t\t\t\t\t<\/figure>\n\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-dbc4739 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"dbc4739\" 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-e900fe4\" data-id=\"e900fe4\" 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-fdc907b elementor-widget elementor-widget-text-editor\" data-id=\"fdc907b\" 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;\">\u2192<\/span><b> VGG16 a donc r\u00e9ussi \u00e0 pr\u00e9dire avec une grande confiance (95,6%) que non seulement l\u2019image donn\u00e9e en entr\u00e9e \u00e9tait un loup mais est m\u00eame all\u00e9 plus loin en pr\u00e9cisant sa race \u00e0 savoir ici un loup gris de l\u2019Est (timber wolf). Impressionnant non\u00a0?<\/b><\/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-9e6adec elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"9e6adec\" 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-9d23601\" data-id=\"9d23601\" 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-3a63a95 elementor-widget elementor-widget-text-editor\" data-id=\"3a63a95\" 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;\">Finalement, le principe de fonctionnement d\u2019un CNN est assez facile \u00e0 comprendre, mais paradoxalement, la mise en place d&rsquo;un tel proc\u00e9d\u00e9 pour classifier des images demeure tr\u00e8s complexe \u00e9tant donn\u00e9 le nombre consid\u00e9rable de param\u00e8tres \u00e0 d\u00e9finir : nombre, taille, d\u00e9placement des filtres, choix de la m\u00e9thode de pooling, choix du nombre de couches de neurones, nombre de neurones par couches, etc.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">Pour pallier \u00e0 cet obstacle, Python \u00e0 travers le module Torchvision, offre la possibilit\u00e9 d\u2019exploiter des mod\u00e8les CNN pr\u00e9-entra\u00een\u00e9s performants tels que VGG16, Resnet101, etc.<\/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-59df323 elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"59df323\" 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-82c0203\" data-id=\"82c0203\" 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-259b7f4 elementor-widget elementor-widget-text-editor\" data-id=\"259b7f4\" 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>Nous avons d\u00e9fini dans cet article le <b>fonctionnement<\/b> et l\u2019<b>architecture des Convolutional Neural Network<\/b>, en nous concentrant sur sa sp\u00e9cificit\u00e9: la partie convolutive.\u00a0<\/p><p>Il nous reste encore \u00e0 appr\u00e9hender une <b>\u00e9tape de la classification<\/b> : la r\u00e9tropropagation du gradient de l\u2019erreur, c\u00e9l\u00e8bre <b>algorithme de descente de gradient<\/b>. Ne vous inqui\u00e9tez pas, un prochain \u00e9pisode sur le sujet arrive bient\u00f4t sur le Blog !<\/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<div class=\"elementor-element elementor-element-daebdd0 e-flex e-con-boxed e-con\" data-id=\"daebdd0\" 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-eaa03fd elementor-widget elementor-widget-text-editor\" data-id=\"eaa03fd\" 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=\"text-decoration: underline;\"><strong>Article 1<\/strong><\/span>\u00a0: <a href=\"https:\/\/datascientest.com\/introduction-deep-learning\">Introduction au deep learning et aux r\u00e9seaux de neurones<\/a><\/p><p><span style=\"text-decoration: underline;\"><strong>Article 2<\/strong><\/span> :\u00a0Convolutional neural network<\/p><p><span style=\"text-decoration: underline;\"><strong>Article 3<\/strong><\/span> : <a href=\"https:\/\/datascientest.com\/fonctionnement-des-reseaux-neurones\">Fonctionnement des r\u00e9seaux de neurones<\/a>\u00a0<\/p><p><span style=\"text-decoration: underline;\"><strong>Article 4<\/strong><\/span> : <a href=\"https:\/\/datascientest.com\/deep_learning_transfer_learning\">Transfer learning<\/a>\u00a0<\/p>\t\t\t\t\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<div class=\"elementor-element elementor-element-135913a e-flex e-con-boxed e-con\" data-id=\"135913a\" 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-fd1c231 elementor-widget elementor-widget-html\" data-id=\"fd1c231\" data-element_type=\"widget\" data-widget_type=\"html.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<script type=\"application\/ld+json\">\r\n{\r\n  \"@context\": \"https:\/\/schema.org\",\r\n  \"@type\": \"FAQPage\",\r\n  \"mainEntity\": [{\r\n    \"@type\": \"Question\",\r\n    \"name\": \"Architecture d\u2019un Convolutional Neural Network-CNN \",\r\n    \"acceptedAnswer\": {\r\n      \"@type\": \"Answer\",\r\n      \"text\": \"Les CNN d\u00e9signent une sous-cat\u00e9gorie de r\u00e9seaux de neurones et sont \u00e0 ce jour un des mod\u00e8les de classification d\u2019images r\u00e9put\u00e9s \u00eatre les plus performant.\"\r\n    }\r\n  },{\r\n    \"@type\": \"Question\",\r\n    \"name\": \"Exemple d\u2019un filtre de convolution classique  \",\r\n    \"acceptedAnswer\": {\r\n      \"@type\": \"Answer\",\r\n      \"text\": \"Lors de la partie convolutive d\u2019un Convolutional Neural Network, l\u2019image fournie en entr\u00e9e passe \u00e0 travers une succession de filtres de convolution. Par exemple, il existe des filtres de convolution fr\u00e9quemment utilis\u00e9s et permettant d\u2019extraire des caract\u00e9ristiques plus pertinentes que des pixels comme la d\u00e9tection des bords (filtre d\u00e9rivateur) ou des formes g\u00e9om\u00e9triques. Le choix et l\u2019application des filtres se fait automatiquement par le mod\u00e8le.\"\r\n    }\r\n  },{\r\n    \"@type\": \"Question\",\r\n    \"name\": \"M\u00e9thode de sous \u00e9chantillonnage : le Max-Pooling\",\r\n    \"acceptedAnswer\": {\r\n      \"@type\": \"Answer\",\r\n      \"text\": \"Le Max-Pooling est un processus de discr\u00e9tisation bas\u00e9 sur des \u00e9chantillons. Son objectif est de sous-\u00e9chantillonner une repr\u00e9sentation d'entr\u00e9e (image, matrice de sortie de couche cach\u00e9e, etc.) en r\u00e9duisant sa dimension. De plus, son int\u00e9r\u00eat est qu\u2019il r\u00e9duit le co\u00fbt de calcul en r\u00e9duisant le nombre de param\u00e8tres \u00e0 apprendre et fournit une invariance par petites translations (si une petite translation ne modifie pas le maximum de la r\u00e9gion balay\u00e9e, le maximum de chaque r\u00e9gion restera le m\u00eame et donc la nouvelle matrice cr\u00e9\u00e9e restera identique).\"\r\n    }\r\n  }]\r\n}\r\n<\/script>\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\t\t\t<\/div>\n\t\t","protected":false},"excerpt":{"rendered":"<p>Dans cette partie, nous allons nous focaliser sur un des algorithmes les plus performants du Deep Learning : les&nbsp;Convolutional Neural Network&nbsp;(CNN) ou r\u00e9seaux de neurones convolutifs en fran\u00e7ais. Ce sont des mod\u00e8les de programmation puissants permettant notamment la reconnaissance d\u2019images en attribuant automatiquement \u00e0 chaque image fournie en entr\u00e9e, une \u00e9tiquette correspondant \u00e0 sa classe [&hellip;]<\/p>\n","protected":false},"author":78,"featured_media":188602,"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>Convolutional Neural Network : Tout ce qu&#039;il y a \u00e0 savoir<\/title>\n<meta name=\"description\" content=\"Nous allons nous focaliser sur un des algorithmes les plus performants du Deep Learning : les\u00a0Convolutional Neural Network\u00a0(CNN) ou r\u00e9seaux de neurones convolutifs en fran\u00e7ais.\" \/>\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\/convolutional-neural-network\" \/>\n<meta property=\"og:locale\" content=\"fr_FR\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Convolutional Neural Network : Tout ce qu&#039;il y a \u00e0 savoir\" \/>\n<meta property=\"og:url\" content=\"https:\/\/datascientest.com\/convolutional-neural-network\" \/>\n<meta property=\"og:site_name\" content=\"Formation Data Science | DataScientest.com\" \/>\n<meta property=\"article:published_time\" content=\"2020-06-25T14:00:33+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2024-01-31T09:31:50+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Illu_BLOG_CNN.png\" \/>\n\t<meta property=\"og:image:width\" content=\"1024\" \/>\n\t<meta property=\"og:image:height\" content=\"563\" \/>\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=\"14 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/datascientest.com\/convolutional-neural-network\",\"url\":\"https:\/\/datascientest.com\/convolutional-neural-network\",\"name\":\"Convolutional Neural Network : Tout ce qu'il y a \u00e0 savoir\",\"isPartOf\":{\"@id\":\"https:\/\/datascientest.com\/#website\"},\"datePublished\":\"2020-06-25T14:00:33+00:00\",\"dateModified\":\"2024-01-31T09:31:50+00:00\",\"author\":{\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/7af2361c48e24e79a8aa2360e6e74925\"},\"description\":\"Nous allons nous focaliser sur un des algorithmes les plus performants du Deep Learning : les\u00a0Convolutional Neural Network\u00a0(CNN) ou r\u00e9seaux de neurones convolutifs en fran\u00e7ais.\",\"inLanguage\":\"fr-FR\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/datascientest.com\/convolutional-neural-network\"]}]},{\"@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":"Convolutional Neural Network : Tout ce qu'il y a \u00e0 savoir","description":"Nous allons nous focaliser sur un des algorithmes les plus performants du Deep Learning : les\u00a0Convolutional Neural Network\u00a0(CNN) ou r\u00e9seaux de neurones convolutifs en fran\u00e7ais.","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\/convolutional-neural-network","og_locale":"fr_FR","og_type":"article","og_title":"Convolutional Neural Network : Tout ce qu'il y a \u00e0 savoir","og_url":"https:\/\/datascientest.com\/convolutional-neural-network","og_site_name":"Formation Data Science | DataScientest.com","article_published_time":"2020-06-25T14:00:33+00:00","article_modified_time":"2024-01-31T09:31:50+00:00","og_image":[{"width":1024,"height":563,"url":"https:\/\/datascientest.com\/wp-content\/uploads\/2020\/06\/Illu_BLOG_CNN.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":"14 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/datascientest.com\/convolutional-neural-network","url":"https:\/\/datascientest.com\/convolutional-neural-network","name":"Convolutional Neural Network : Tout ce qu'il y a \u00e0 savoir","isPartOf":{"@id":"https:\/\/datascientest.com\/#website"},"datePublished":"2020-06-25T14:00:33+00:00","dateModified":"2024-01-31T09:31:50+00:00","author":{"@id":"https:\/\/datascientest.com\/#\/schema\/person\/7af2361c48e24e79a8aa2360e6e74925"},"description":"Nous allons nous focaliser sur un des algorithmes les plus performants du Deep Learning : les\u00a0Convolutional Neural Network\u00a0(CNN) ou r\u00e9seaux de neurones convolutifs en fran\u00e7ais.","inLanguage":"fr-FR","potentialAction":[{"@type":"ReadAction","target":["https:\/\/datascientest.com\/convolutional-neural-network"]}]},{"@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\/4423"}],"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=4423"}],"version-history":[{"count":0,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts\/4423\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/media\/188602"}],"wp:attachment":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/media?parent=4423"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/categories?post=4423"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/tags?post=4423"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}