{"id":72795,"date":"2021-06-28T08:56:00","date_gmt":"2021-06-28T07:56:00","guid":{"rendered":"https:\/\/datascientest.com\/?p=72795"},"modified":"2023-11-09T14:06:58","modified_gmt":"2023-11-09T14:06:58","slug":"reseaux-de-neurones-densenet","status":"publish","type":"post","link":"https:\/\/datascientest.com\/reseaux-de-neurones-densenet","title":{"rendered":"R\u00e9seaux de neurones DenseNet : tout ce qu&rsquo;il y a \u00e0 savoir"},"content":{"rendered":"\t\t<div data-elementor-type=\"wp-post\" data-elementor-id=\"72795\" class=\"elementor elementor-72795\" data-elementor-post-type=\"post\">\n\t\t\t\t\t\t\t\t\t<section class=\"elementor-section elementor-top-section elementor-element elementor-element-7bd29fd elementor-section-boxed elementor-section-height-default elementor-section-height-default\" data-id=\"7bd29fd\" 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-d6440ae\" data-id=\"d6440ae\" 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-c558cc0 elementor-widget elementor-widget-heading\" data-id=\"c558cc0\" 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\">Vous avez d\u00e9j\u00e0 probablement entendu parler des r\u00e9seaux convolutifs dans les contextes de reconnaissance d\u2019image, de traitement vid\u00e9o ou encore dans le ciblage publicitaire et le NLP. Cet article a pour but de d\u00e9livrer une premi\u00e8re approche analytique et une compr\u00e9hension de l\u2019architecture de ces r\u00e9seaux, et plus particuli\u00e8rement sur l\u2019architecture des r\u00e9seaux DenseNets.<\/h2>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-e7dedfb elementor-widget elementor-widget-heading\" data-id=\"e7dedfb\" 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\">Les r\u00e9seaux de neurones convolutifs - CNN<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-416bb0f elementor-widget elementor-widget-text-editor\" data-id=\"416bb0f\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<style>\/*! elementor - v3.15.0 - 02-08-2023 *\/\n.elementor-widget-text-editor.elementor-drop-cap-view-stacked .elementor-drop-cap{background-color:#69727d;color:#fff}.elementor-widget-text-editor.elementor-drop-cap-view-framed .elementor-drop-cap{color:#69727d;border:3px solid;background-color:transparent}.elementor-widget-text-editor:not(.elementor-drop-cap-view-default) .elementor-drop-cap{margin-top:8px}.elementor-widget-text-editor:not(.elementor-drop-cap-view-default) .elementor-drop-cap-letter{width:1em;height:1em}.elementor-widget-text-editor .elementor-drop-cap{float:left;text-align:center;line-height:1;font-size:50px}.elementor-widget-text-editor .elementor-drop-cap-letter{display:inline-block}<\/style>\t\t\t\t<p><span style=\"font-weight: 400;\">Prenons le cas de l\u2019analyse d\u2019image qui fait partie aujourd\u2019hui d\u2019une des grandes applications du machine learning. <\/span><span style=\"font-weight: 400;\">Chaque image est repr\u00e9sent\u00e9e par un tr\u00e8s grand nombre de donn\u00e9es et peut \u00eatre visualis\u00e9e comme un objet au sein d\u2019un espace de tr\u00e8s grande dimension.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">Mais alors <\/span><b>comment exploiter les propri\u00e9t\u00e9s de ces images<\/b><span style=\"font-weight: 400;\"> de fa\u00e7on efficace compte tenu de la taille cons\u00e9quentes des donn\u00e9es qui les repr\u00e9sentent\u202f?<br \/><br \/><\/span><span style=\"font-weight: 400;\">Les r\u00e9seaux de convolution apportent une solution efficace \u00e0 ce probl\u00e8me : chaque couche interm\u00e9diaire du r\u00e9seau est expos\u00e9e \u00e0 une portion de l\u2019image appel\u00e9e<\/span> <b><i>champ r\u00e9cepteur<\/i><\/b><span style=\"font-weight: 400;\"> pour lequel plusieurs neurones effectuant diff\u00e9rentes actions, calculent une <\/span><b>matrice de convolution <\/b><span style=\"font-weight: 400;\">correspondante. <\/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-9d2a11e elementor-widget elementor-widget-image\" data-id=\"9d2a11e\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t<style>\/*! elementor - v3.15.0 - 02-08-2023 *\/\n.elementor-widget-image{text-align:center}.elementor-widget-image a{display:inline-block}.elementor-widget-image a img[src$=\".svg\"]{width:48px}.elementor-widget-image img{vertical-align:middle;display:inline-block}<\/style>\t\t\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"496\" height=\"274\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2021\/06\/pasted-image-0.png\" class=\"attachment-large size-large wp-image-79533\" alt=\"\" loading=\"lazy\" \/>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-f36dce7 elementor-align-center button-principale elementor-widget elementor-widget-button\" data-id=\"f36dce7\" 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\">En savoir plus sur la 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<div class=\"elementor-element elementor-element-3c4a514 elementor-widget elementor-widget-text-editor\" data-id=\"3c4a514\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Ainsi, chaque couche interm\u00e9diaire g\u00e9n\u00e8re plusieurs matrices de convolution : tout cela laisse penser que nous \u00e9tendons l\u2019information plut\u00f4t que de la restreindre ? Pas vraiment\u202f!<\/span><\/p><p><span style=\"font-weight: 400;\">Le partage de <\/span><b>poids synaptique<\/b><span style=\"font-weight: 400;\"> (ou <\/span><i><span style=\"font-weight: 400;\">weight sharing<\/span><\/i><span style=\"font-weight: 400;\">), d\u00e9finit le partage de poids entre les neurones d\u2019une m\u00eame couche. <\/span><span style=\"font-weight: 400;\">Ceci correspond \u00e0 une <\/span><b>r\u00e9duction consid\u00e9rable des param\u00e8tres<\/b><span style=\"font-weight: 400;\"> au sein de chaque couche. L\u2019information \u00e0 la sortie de ces couches est \u00e9galement restreinte gr\u00e2ce \u00e0 une phase de<\/span><b> r\u00e9duction de dimension<\/b><span style=\"font-weight: 400;\"> qui permet de rassembler les informations\u00a0 provenant de plusieurs neurones voisins appartenant au m\u00eame canal, en une information commune au travers de l\u2019ajout syst\u00e9matique d\u2019une couche d\u2019extraction en sortie des couches convolutives.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">Il existe deux sortes de couche d\u2019extraction :<\/span><\/p><ul><li style=\"font-weight: 400;\" aria-level=\"1\"><strong>Average pooling :<\/strong><span style=\"font-weight: 400;\"> qui calcule la valeur moyenne des informations<\/span><\/li><li style=\"font-weight: 400;\" aria-level=\"1\"><strong>Max pooling :<\/strong><span style=\"font-weight: 400;\"> qui ne retient que la valeur maximale des informations <\/span>et les utilise pour cr\u00e9er une <i>feature map<\/i> sous \u00e9chantillonn\u00e9e.<\/li><\/ul>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-129500f elementor-widget elementor-widget-image\" data-id=\"129500f\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"300\" height=\"262\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2021\/05\/yLDAvFWcjzR0Pf1Oe5BEXXHkgVMzq82H-5I7sYCaqryoDyTloX7ioq8xQymIiBtq9pSHbqL5ZuLG2qpK1qb4iyC_4_bPZEO4DLPGAY6cy-hVbwllUkHVdYO5x0Z0TQlph-300x262.png\" class=\"attachment-medium size-medium wp-image-72798\" alt=\"couches d\u2019extraction\" loading=\"lazy\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2021\/05\/yLDAvFWcjzR0Pf1Oe5BEXXHkgVMzq82H-5I7sYCaqryoDyTloX7ioq8xQymIiBtq9pSHbqL5ZuLG2qpK1qb4iyC_4_bPZEO4DLPGAY6cy-hVbwllUkHVdYO5x0Z0TQlph-300x262.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2021\/05\/yLDAvFWcjzR0Pf1Oe5BEXXHkgVMzq82H-5I7sYCaqryoDyTloX7ioq8xQymIiBtq9pSHbqL5ZuLG2qpK1qb4iyC_4_bPZEO4DLPGAY6cy-hVbwllUkHVdYO5x0Z0TQlph.png 569w\" sizes=\"(max-width: 300px) 100vw, 300px\" \/>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-2e45c02 elementor-widget elementor-widget-text-editor\" data-id=\"2e45c02\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Nous obtenons donc en sortie de ces superpositions (couche concolutives et couche avg\/max pooling), un vecteur de taille raisonnable et r\u00e9duite repr\u00e9sentant les informations extraites de l\u2019image de fa\u00e7on certes, moins pr\u00e9cise, mais beaucoup plus facilement exploitables.<\/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-9999cba elementor-widget elementor-widget-image\" data-id=\"9999cba\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"1024\" height=\"562\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2021\/06\/illu_DenseNet-09.png\" class=\"attachment-full size-full wp-image-77955\" alt=\"\" loading=\"lazy\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2021\/06\/illu_DenseNet-09.png 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2021\/06\/illu_DenseNet-09-300x165.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2021\/06\/illu_DenseNet-09-768x422.png 768w\" sizes=\"(max-width: 1024px) 100vw, 1024px\" \/>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-6d4262d elementor-widget elementor-widget-heading\" data-id=\"6d4262d\" 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\">Les r\u00e9seaux DenseNet<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-d64d029 elementor-widget elementor-widget-text-editor\" data-id=\"d64d029\" 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>Les <a href=\"https:\/\/towardsdatascience.com\/review-densenet-image-classification-b6631a8ef803\">r\u00e9seaux DenseNet<\/a> sont des <a href=\"https:\/\/datascientest.com\/convolutional-neural-network\">r\u00e9seaux de neurones convolutifs<\/a> r\u00e9pondant \u00e0 une certaine architecture commun\u00e9ment appel\u00e9e <b>r\u00e9seaux de neurones dens\u00e9ment connect\u00e9s<\/b><span style=\"font-weight: 400;\"> (\u2018<\/span><i><span style=\"font-weight: 400;\">densely connected convolutional network\u2019<\/span><\/i><span style=\"font-weight: 400;\"> ou DenseNet).<\/span><\/p><p><span style=\"font-weight: 400;\">Un challenge grandissant de ce type de r\u00e9seau est l&rsquo;<\/span><b>optimisation des performances<\/b><span style=\"font-weight: 400;\">. <\/span><span style=\"font-weight: 400;\">Une solution na\u00efve serait simplement d\u2019empiler davantage de couches convolutives: un probl\u00e8me \u00e9vident qui survient d\u00e8s lors, est l\u2019agrandissement de la profondeur du r\u00e9seau. <\/span><span style=\"font-weight: 400;\">Lorsque l\u2019on effectue la r\u00e9tropropagation du gradient au travers des couches: on effectue en r\u00e9alit\u00e9 une op\u00e9ration sur les d\u00e9riv\u00e9es partielles de chaque couche cach\u00e9e du r\u00e9seau et cela peut potentiellement compliquer les mises \u00e0 jour des poids durant la phase d&rsquo;entra\u00eenement.<\/span><\/p><p><span style=\"font-weight: 400;\">La grande sp\u00e9cificit\u00e9 de cette architecture r\u00e9side dans l&rsquo;entr\u00e9e des couches qui rassemble par concat\u00e9nation toutes les entr\u00e9es des couches pr\u00e9c\u00e9dentes (voir sch\u00e9ma) : cr\u00e9ant ainsi une <\/span><i><span style=\"font-weight: 400;\">feature map,<\/span><\/i><span style=\"font-weight: 400;\"> tout en conservant une r\u00e9solution spatiale identique : on parle de <\/span><b><i>channel-wise concatenation<\/i><\/b><span style=\"font-weight: 400;\">.<\/span><\/p>\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-d0264cf elementor-widget elementor-widget-image\" data-id=\"d0264cf\" data-element_type=\"widget\" data-widget_type=\"image.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<img decoding=\"async\" width=\"1024\" height=\"562\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2021\/06\/illu_DenseNet_schema_2.png\" class=\"attachment-full size-full wp-image-77956\" alt=\"\" loading=\"lazy\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2021\/06\/illu_DenseNet_schema_2.png 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2021\/06\/illu_DenseNet_schema_2-300x165.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2021\/06\/illu_DenseNet_schema_2-768x422.png 768w\" sizes=\"(max-width: 1024px) 100vw, 1024px\" \/>\t\t\t\t\t\t\t\t\t\t\t\t\t\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-b92c37c elementor-align-center button-principale elementor-widget elementor-widget-button\" data-id=\"b92c37c\" 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 le parcours 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<div class=\"elementor-element elementor-element-1945eff elementor-widget elementor-widget-text-editor\" data-id=\"1945eff\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Ainsi, il devient inutile d\u2019effectuer un processus de s\u00e9lection d\u2019information, contrairement \u00e0 d&rsquo;autres r\u00e9seaux pour lesquels il est <strong>n\u00e9cessaire de choisir quelles informations<\/strong> sont \u00e0 transmettre aux couches suivantes.<\/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-4a1bf6f elementor-widget elementor-widget-image\" data-id=\"4a1bf6f\" 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=\"1024\" height=\"562\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2021\/06\/illu_densenet_schema_3-20-1.png\" class=\"attachment-full size-full wp-image-78982\" alt=\"\" loading=\"lazy\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2021\/06\/illu_densenet_schema_3-20-1.png 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2021\/06\/illu_densenet_schema_3-20-1-300x165.png 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2021\/06\/illu_densenet_schema_3-20-1-768x422.png 768w\" sizes=\"(max-width: 1024px) 100vw, 1024px\" \/>\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">Assemblage de plusieurs blocs Dense avec couches de transition<\/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-8fc061d elementor-widget elementor-widget-text-editor\" data-id=\"8fc061d\" 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\u2019addition des <i>dense blocs<\/i><\/span><span style=\"font-weight: 400;\">\u00a0rend le r\u00e9seau profond et peut parfois s&rsquo;av\u00e9rer probl\u00e9matique. Ceci peut \u00eatre solutionn\u00e9 par la scission de ces blocs et par l\u2019introduction de couches de convolution 1-by-1 qui conservent la r\u00e9solution spatiale tout en r\u00e9duisant la profondeur de la<\/span><i><span style=\"font-weight: 400;\"> feature map<\/span><\/i><span style=\"font-weight: 400;\">.<\/span><\/p><p><span style=\"font-weight: 400;\">Comme explicit\u00e9 dans la partie pr\u00e9c\u00e9dente, une couche de Pooling (MaxPool ou AvgPool) est souvent empil\u00e9e \u00e0 la sortie de ces couches dans le but de r\u00e9duire la dimension de la <\/span><i><span style=\"font-weight: 400;\">feature map <\/span><\/i><span style=\"font-weight: 400;\">et de faciliter l\u2019exploitation des donn\u00e9es.<\/span><\/p><p><span style=\"font-weight: 400;\">Il est important de noter que les feature map sont de m\u00eame taille \u00e0 l&rsquo;int\u00e9rieur de chaque bloc, ce qui rend la concatenation moins complexe.<\/span><span style=\"font-weight: 400;\"><br \/><\/span><span style=\"font-weight: 400;\"><br \/><\/span><span style=\"font-weight: 400;\">Parmi les avantages du r\u00e9seau DenseNet, nous pouvons mentionner la\u00a0 <\/span><b>r\u00e9tropropagation facilit\u00e9e du signal d\u2019erreur<\/b><span style=\"font-weight: 400;\"> de mani\u00e8re directe. Nous pouvons \u00e9galement mettre en \u00e9vidence l&rsquo;<\/span><b>efficacit\u00e9 computationnelle de ce syst\u00e8me <\/b><span style=\"font-weight: 400;\">notamment gr\u00e2ce \u00e0 la gestion du nombre de param\u00e8tres, DenseNet assure la diversit\u00e9 des features en comparaison avec un r\u00e9seau convolutif traditionnel ainsi que leur faible complexit\u00e9 ce qui permet des performances optimales.<\/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-0b33444 elementor-widget elementor-widget-text-editor\" data-id=\"0b33444\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">En conclusion, l\u2019architecture de r\u00e9seau DenseNet s\u2019inscrit dans la famille des r\u00e9seaux de neurones convolutifs et permet une optimisation des performances notamment dans les domaines de traitement d\u2019images.\u00a0<\/span><\/p><p><span style=\"font-weight: 400;\">Sa structure singuli\u00e8re se compose d\u2019un empilement de <\/span><i><span style=\"font-weight: 400;\">dense block<\/span><\/i><span style=\"font-weight: 400;\">, de couches de convolution et de pooling successivement. Les couches de Pooling (average ou max) permettent de garantir une restriction de la dimension des donn\u00e9es ce qui justifie l&rsquo;efficacit\u00e9 de traitement de ce type de r\u00e9seau.<\/span><\/p><p><span style=\"font-weight: 400;\">De la reconnaissance faciale jusqu\u2019au traitement de langues naturelles, les applications du r\u00e9seau convolutif DenseNet \u00e0 l\u2019AI en sont tr\u00e8s larges.<\/span><\/p><p><span style=\"font-weight: 400;\">Si vous voulez en savoir plus sur ces sujets d\u2019applications, vous pouvez lire notre article sur le <a href=\"https:\/\/datascientest.com\/nlp-twitter-analyse-de-sentiment\">NLP et l\u2019analyse de sentiments<\/a>ou alors d\u00e9couvrir nos diff\u00e9rents cursus de formation pour apprendre \u00e0 maitriser ces sujets.<\/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-597c947 elementor-align-center button-principale elementor-widget elementor-widget-button\" data-id=\"597c947\" 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\/formations-data\">\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 nos formations en Data Science<\/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\t\t\t<\/div>\n\t\t","protected":false},"excerpt":{"rendered":"<p>Vous avez d\u00e9j\u00e0 probablement entendu parler des r\u00e9seaux convolutifs dans les contextes de reconnaissance d\u2019image, de traitement vid\u00e9o ou encore dans le ciblage publicitaire et le NLP. Cet article a pour but de d\u00e9livrer une premi\u00e8re approche analytique et une compr\u00e9hension de l\u2019architecture de ces r\u00e9seaux, et plus particuli\u00e8rement sur l\u2019architecture des r\u00e9seaux DenseNets. Les [&hellip;]<\/p>\n","protected":false},"author":77,"featured_media":78981,"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>R\u00e9seaux de neurones DenseNet : tout ce qu&#039;il y a \u00e0 savoir<\/title>\n<meta name=\"description\" content=\"Vous avez d\u00e9j\u00e0 probablement entendu parler des r\u00e9seaux convolutifs dans les contextes de reconnaissance d\u2019image, de traitement vid\u00e9o ou encore dans le ciblage publicitaire et le NLP. Cet article a pour but de d\u00e9livrer une premi\u00e8re approche analytique et une compr\u00e9hension de l\u2019architecture de ces r\u00e9seaux, et plus particuli\u00e8rement sur l\u2019architecture des r\u00e9seaux DenseNets.\" \/>\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\/reseaux-de-neurones-densenet\" \/>\n<meta property=\"og:locale\" content=\"fr_FR\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"R\u00e9seaux de neurones DenseNet : tout ce qu&#039;il y a \u00e0 savoir\" \/>\n<meta property=\"og:url\" content=\"https:\/\/datascientest.com\/reseaux-de-neurones-densenet\" \/>\n<meta property=\"og:site_name\" content=\"Formation Data Science | DataScientest.com\" \/>\n<meta property=\"article:published_time\" content=\"2021-06-28T07:56:00+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2023-11-09T14:06:58+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/datascientest.com\/wp-content\/uploads\/2021\/06\/illu_DenseNet_schema_2-1.png\" \/>\n\t<meta property=\"og:image:width\" content=\"1024\" \/>\n\t<meta property=\"og:image:height\" content=\"562\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/png\" \/>\n<meta name=\"author\" content=\"Raphael Kassel\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"\u00c9crit par\" \/>\n\t<meta name=\"twitter:data1\" content=\"Raphael Kassel\" \/>\n\t<meta name=\"twitter:label2\" content=\"Dur\u00e9e de lecture estim\u00e9e\" \/>\n\t<meta name=\"twitter:data2\" content=\"6 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/datascientest.com\/reseaux-de-neurones-densenet\",\"url\":\"https:\/\/datascientest.com\/reseaux-de-neurones-densenet\",\"name\":\"R\u00e9seaux de neurones DenseNet : tout ce qu'il y a \u00e0 savoir\",\"isPartOf\":{\"@id\":\"https:\/\/datascientest.com\/#website\"},\"datePublished\":\"2021-06-28T07:56:00+00:00\",\"dateModified\":\"2023-11-09T14:06:58+00:00\",\"author\":{\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/6eca16f63daaafcbc96e15549b9d4ee1\"},\"description\":\"Vous avez d\u00e9j\u00e0 probablement entendu parler des r\u00e9seaux convolutifs dans les contextes de reconnaissance d\u2019image, de traitement vid\u00e9o ou encore dans le ciblage publicitaire et le NLP. Cet article a pour but de d\u00e9livrer une premi\u00e8re approche analytique et une compr\u00e9hension de l\u2019architecture de ces r\u00e9seaux, et plus particuli\u00e8rement sur l\u2019architecture des r\u00e9seaux DenseNets.\",\"inLanguage\":\"fr-FR\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/datascientest.com\/reseaux-de-neurones-densenet\"]}]},{\"@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\/6eca16f63daaafcbc96e15549b9d4ee1\",\"name\":\"Raphael Kassel\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"fr-FR\",\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/raph.jpeg\",\"contentUrl\":\"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/raph.jpeg\",\"caption\":\"Raphael Kassel\"},\"description\":\"Raphael Kassel est le Head of Data Programs chez DataScientest. Il supervise une \u00e9quipe Data, impl\u00e9mente de nouveaux parcours comme \\\"Chef de projet IA\\\" et \\\"Data Product Manager\\\", et contribue \u00e0 la modernisation des op\u00e9rations du groupe. En collaboration \u00e9troite avec l'\u00e9quipe de d\u00e9veloppement, il adapte l'accompagnement acad\u00e9mique aux \u00e9volutions de l'\u00e9quipe data et \u00e0 l'internationalisation.\",\"sameAs\":[\"https:\/\/www.linkedin.com\/in\/raphaelkassel\/\",\"Data Science\"],\"url\":\"https:\/\/datascientest.com\/author\/raphaelkassel\"}]}<\/script>\n<!-- \/ Yoast SEO Premium plugin. -->","yoast_head_json":{"title":"R\u00e9seaux de neurones DenseNet : tout ce qu'il y a \u00e0 savoir","description":"Vous avez d\u00e9j\u00e0 probablement entendu parler des r\u00e9seaux convolutifs dans les contextes de reconnaissance d\u2019image, de traitement vid\u00e9o ou encore dans le ciblage publicitaire et le NLP. Cet article a pour but de d\u00e9livrer une premi\u00e8re approche analytique et une compr\u00e9hension de l\u2019architecture de ces r\u00e9seaux, et plus particuli\u00e8rement sur l\u2019architecture des r\u00e9seaux DenseNets.","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\/reseaux-de-neurones-densenet","og_locale":"fr_FR","og_type":"article","og_title":"R\u00e9seaux de neurones DenseNet : tout ce qu'il y a \u00e0 savoir","og_url":"https:\/\/datascientest.com\/reseaux-de-neurones-densenet","og_site_name":"Formation Data Science | DataScientest.com","article_published_time":"2021-06-28T07:56:00+00:00","article_modified_time":"2023-11-09T14:06:58+00:00","og_image":[{"width":1024,"height":562,"url":"https:\/\/datascientest.com\/wp-content\/uploads\/2021\/06\/illu_DenseNet_schema_2-1.png","type":"image\/png"}],"author":"Raphael Kassel","twitter_card":"summary_large_image","twitter_misc":{"\u00c9crit par":"Raphael Kassel","Dur\u00e9e de lecture estim\u00e9e":"6 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/datascientest.com\/reseaux-de-neurones-densenet","url":"https:\/\/datascientest.com\/reseaux-de-neurones-densenet","name":"R\u00e9seaux de neurones DenseNet : tout ce qu'il y a \u00e0 savoir","isPartOf":{"@id":"https:\/\/datascientest.com\/#website"},"datePublished":"2021-06-28T07:56:00+00:00","dateModified":"2023-11-09T14:06:58+00:00","author":{"@id":"https:\/\/datascientest.com\/#\/schema\/person\/6eca16f63daaafcbc96e15549b9d4ee1"},"description":"Vous avez d\u00e9j\u00e0 probablement entendu parler des r\u00e9seaux convolutifs dans les contextes de reconnaissance d\u2019image, de traitement vid\u00e9o ou encore dans le ciblage publicitaire et le NLP. Cet article a pour but de d\u00e9livrer une premi\u00e8re approche analytique et une compr\u00e9hension de l\u2019architecture de ces r\u00e9seaux, et plus particuli\u00e8rement sur l\u2019architecture des r\u00e9seaux DenseNets.","inLanguage":"fr-FR","potentialAction":[{"@type":"ReadAction","target":["https:\/\/datascientest.com\/reseaux-de-neurones-densenet"]}]},{"@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\/6eca16f63daaafcbc96e15549b9d4ee1","name":"Raphael Kassel","image":{"@type":"ImageObject","inLanguage":"fr-FR","@id":"https:\/\/datascientest.com\/#\/schema\/person\/image\/","url":"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/raph.jpeg","contentUrl":"https:\/\/datascientest.com\/wp-content\/uploads\/2023\/11\/raph.jpeg","caption":"Raphael Kassel"},"description":"Raphael Kassel est le Head of Data Programs chez DataScientest. Il supervise une \u00e9quipe Data, impl\u00e9mente de nouveaux parcours comme \"Chef de projet IA\" et \"Data Product Manager\", et contribue \u00e0 la modernisation des op\u00e9rations du groupe. En collaboration \u00e9troite avec l'\u00e9quipe de d\u00e9veloppement, il adapte l'accompagnement acad\u00e9mique aux \u00e9volutions de l'\u00e9quipe data et \u00e0 l'internationalisation.","sameAs":["https:\/\/www.linkedin.com\/in\/raphaelkassel\/","Data Science"],"url":"https:\/\/datascientest.com\/author\/raphaelkassel"}]}},"_links":{"self":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts\/72795"}],"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\/77"}],"replies":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/comments?post=72795"}],"version-history":[{"count":0,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts\/72795\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/media\/78981"}],"wp:attachment":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/media?parent=72795"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/categories?post=72795"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/tags?post=72795"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}