{"id":176176,"date":"2022-10-14T09:46:44","date_gmt":"2022-10-14T09:46:44","guid":{"rendered":"https:\/\/datascientest.com\/?p=176176"},"modified":"2023-11-09T13:56:11","modified_gmt":"2023-11-09T13:56:11","slug":"optical-character-recognition","status":"publish","type":"post","link":"https:\/\/datascientest.com\/optical-character-recognition","title":{"rendered":"Apprendre \u00e0 l\u2019ordinateur \u00e0 lire avec l\u2019Optical Character Recognition"},"content":{"rendered":"\t\t<div data-elementor-type=\"wp-post\" data-elementor-id=\"176176\" class=\"elementor elementor-176176\" data-elementor-post-type=\"post\">\n\t\t\t\t\t\t\t<div class=\"elementor-element elementor-element-0e8c4be e-flex e-con-boxed e-con\" data-id=\"0e8c4be\" 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-1b30e7e elementor-widget elementor-widget-text-editor\" data-id=\"1b30e7e\" 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<h2>La reconnaissance optique de caract\u00e8res (ROC ou Optical Character Recognition en anglais), appel\u00e9e \u00e9galement oc\u00e9risation, regroupe l\u2019ensemble des m\u00e9thodes qui permettent de g\u00e9n\u00e9rer des fichiers textes \u00e0 partir d\u2019images contenant du texte manuscrit. Avec l\u2019av\u00e8nement du num\u00e9rique et de l\u2019automatisation, l&rsquo;oc\u00e9risation est devenue un outil incontournable car les images comportant du texte ne sont pas exploitables par un ordinateur.<\/h2>\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-f2bcce9 e-flex e-con-boxed e-con\" data-id=\"f2bcce9\" 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-9fdd625 elementor-widget elementor-widget-heading\" data-id=\"9fdd625\" 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><h3 class=\"elementor-heading-title elementor-size-default\">1. Un peu d\u2019histoire<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-d5c345f elementor-widget elementor-widget-text-editor\" data-id=\"d5c345f\" 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 premi\u00e8re version d\u2019une machine capable de reconna\u00eetre des caract\u00e8res dans un document manuscrit fut <strong>mise au point par<\/strong> <strong>Gustav Tauschek en 1929<\/strong>, un ing\u00e9nieur allemand. Le principe \u00e9tait le suivant. Un document contenant un texte manuscrit \u00e9tait positionn\u00e9 devant la fen\u00eatre de la machine. Ensuite, pour d\u00e9tecter le bon caract\u00e8re, une roue comportant <strong>des trous de la forme des lettres tournaient jusqu\u2019\u00e0 ce que l\u2019image du caract\u00e8re et le trou co\u00efncident parfaitement<\/strong> gr\u00e2ce \u00e0 un photod\u00e9tecteur. Une fois le caract\u00e8re d\u00e9tect\u00e9, il \u00e9tait \u00e9crit une feuille de papier.\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-6a405c4 elementor-widget elementor-widget-image\" data-id=\"6a405c4\" 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<img decoding=\"async\" width=\"442\" height=\"303\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2022\/09\/ocr.png\" class=\"attachment-large size-large wp-image-176179\" alt=\"\" loading=\"lazy\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2022\/09\/ocr.png 442w, https:\/\/datascientest.com\/wp-content\/uploads\/2022\/09\/ocr-300x206.png 300w\" sizes=\"(max-width: 442px) 100vw, 442px\" \/>\t\t\t\t\t\t\t\t\t\t\t<figcaption class=\"widget-image-caption wp-caption-text\">Machine \u00e0 lire de Gustav Tauschek<\/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\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-f064eb2 e-flex e-con-boxed e-con\" data-id=\"f064eb2\" 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-01d419f elementor-widget elementor-widget-heading\" data-id=\"01d419f\" 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\">2. Utilisation du Deep Learning pour l\u2019oc\u00e9risation<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-495aed5 elementor-widget elementor-widget-text-editor\" data-id=\"495aed5\" 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 facilement constater que faire de l\u2019<strong>oc\u00e9risation<\/strong> \u00e0 l\u2019aide de machines m\u00e9caniques, aussi performantes qu\u2019elles soient, peut s\u2019av\u00e9rer tr\u00e8s co\u00fbteux et assez long. Les entreprises qui souhaitent automatiser leur <strong>pipeline de digitalisation<\/strong> ont besoin d\u2019un outil rapide et efficace de <strong>reconnaissance d\u2019\u00e9criture manuscrite<\/strong>. Gr\u00e2ce au d\u00e9veloppement du <a href=\"https:\/\/datascientest.com\/deep-learning\">Deep Learning<\/a> notamment dans la domaine du traitement de l\u2019image, il existe aujourd\u2019hui des <a href=\"https:\/\/datascientest.com\/algorithme\">algorithmes<\/a> capables de r\u00e9soudre des probl\u00e8mes d\u2019oc\u00e9risation. Le mod\u00e8le OCR se d\u00e9compose en 2 \u00e9tapes : la d\u00e9tection et la reconnaissance du texte.<\/span><\/p><ul><li><span style=\"font-weight: 400;\">La premi\u00e8re est le processus qui permet de <strong>d\u00e9tecter les zones de textes dans un document<\/strong> et peut s\u2019av\u00e9rer \u00eatre une t\u00e2che vraiment compliqu\u00e9e car les documents peuvent avoir une structure tr\u00e8s diff\u00e9rente. Les zones de textes ne seront pas au m\u00eame endroit entre un roman et un journal par exemple.\u00a0<\/span><\/li><li><span style=\"font-weight: 400;\">La deuxi\u00e8me permet de <strong>reconna\u00eetre le texte<\/strong> pour une zone du document en contenant.<\/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\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-5283cd4 e-flex e-con-boxed e-con\" data-id=\"5283cd4\" 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-4456e5b elementor-widget elementor-widget-heading\" data-id=\"4456e5b\" 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\">3. D\u00e9tection de texte \u00e0 l\u2019aide du Deep Learning<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-918e1fd elementor-widget elementor-widget-text-editor\" data-id=\"918e1fd\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Il existe plusieurs m\u00e9thodes pour d\u00e9tecter du texte au sein d\u2019un document. Certaines d\u2019entre elles s\u2019inspirent de mod\u00e8les de d\u00e9tection d&rsquo;objets (<strong>Single Shot Box Detection, Faster-RCNN<\/strong>) qui sont utilis\u00e9s pour d\u00e9tecter des visages, des voitures, etc. dans une image. Ce type de mod\u00e8le renvoie une <strong>Bounding Box<\/strong> qui n\u2019est autre qu\u2019un cadre qui entoure l\u2019objet \u00e0 d\u00e9tecter. Pour d\u00e9tecter du texte, ces mod\u00e8les sont modifi\u00e9s et adapt\u00e9s. On pourra citer le mod\u00e8le <strong>TextBoxes<\/strong> qui s\u2019appuie sur le mod\u00e8le <strong>SSD (Single Short Detector)<\/strong> tout en ajoutant des Bounding Boxes plus sp\u00e9cifiques \u00e0 la <strong>d\u00e9tection de texte<\/strong>. Le mod\u00e8le renvoie les zones de l\u2019image contenant du texte.<\/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-759d304 elementor-widget elementor-widget-image\" data-id=\"759d304\" 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.google.com\/url?q=https:\/\/www.researchgate.net\/figure\/The-architecture-of-TextBoxes-a-29-layer-fully-convolutional-network-including-13_fig3_322355323&#038;sa=D&#038;source=docs&#038;ust=1663177989034716&#038;usg=AOvVaw2f0LVJREyJPzCjZ3-7EQOi\">\n\t\t\t\t\t\t\t<img decoding=\"async\" width=\"800\" height=\"241\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2022\/10\/Ex-architecture-modele-reconnaissance-texte-OCR-1-1024x308.jpg\" class=\"attachment-large size-large wp-image-178390\" alt=\"architecture modele de reconnaissance de texte\" loading=\"lazy\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2022\/10\/Ex-architecture-modele-reconnaissance-texte-OCR-1-1024x308.jpg 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2022\/10\/Ex-architecture-modele-reconnaissance-texte-OCR-1-300x90.jpg 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2022\/10\/Ex-architecture-modele-reconnaissance-texte-OCR-1-768x231.jpg 768w, https:\/\/datascientest.com\/wp-content\/uploads\/2022\/10\/Ex-architecture-modele-reconnaissance-texte-OCR-1-1536x463.jpg 1536w, https:\/\/datascientest.com\/wp-content\/uploads\/2022\/10\/Ex-architecture-modele-reconnaissance-texte-OCR-1-2048x617.jpg 2048w\" 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<figcaption class=\"widget-image-caption wp-caption-text\">Exemple d\u2019architecture d\u2019un mod\u00e8le de reconnaissance texte<\/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-12e1763 elementor-widget elementor-widget-text-editor\" data-id=\"12e1763\" data-element_type=\"widget\" data-widget_type=\"text-editor.default\">\n\t\t\t\t<div class=\"elementor-widget-container\">\n\t\t\t\t\t\t\t<p><span style=\"font-weight: 400;\">Il existe aussi d\u2019autres m\u00e9thodes de d\u00e9tection bas\u00e9es sur la <strong>segmentation<\/strong>. La segmentation est un proc\u00e9d\u00e9 qui permet de <strong>d\u00e9composer une image en plusieurs zones<\/strong> appel\u00e9es classes qui comportent des \u00e9l\u00e9ments similaires. En voici un exemple :<\/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-44e5aae elementor-widget elementor-widget-image\" data-id=\"44e5aae\" 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.google.com\/url?q=https:\/\/makina-corpus.com\/societe\/partenariat-avec-lecole-polytechnique-federale-de-lausanne-pour-developper-un-outil-de&#038;sa=D&#038;source=docs&#038;ust=1663177989031206&#038;usg=AOvVaw3mO5Xe4eQBwjiEwbJYjIPo\">\n\t\t\t\t\t\t\t<img decoding=\"async\" width=\"675\" height=\"175\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2022\/09\/ocr2.png\" class=\"attachment-large size-large wp-image-176199\" alt=\"\" loading=\"lazy\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2022\/09\/ocr2.png 675w, https:\/\/datascientest.com\/wp-content\/uploads\/2022\/09\/ocr2-300x78.png 300w\" sizes=\"(max-width: 675px) 100vw, 675px\" \/>\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 de segmentation d\u2019une image en plusieurs classes (voiture, trottoir, personnes, arbres, \u2026)<\/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-6df0896 elementor-widget elementor-widget-text-editor\" data-id=\"6df0896\" 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 la segmentation de texte, on utilise souvent des <strong>r\u00e9seaux de neurones convolutifs (FCN)<\/strong> ainsi que des <strong>r\u00e9seaux de neurones pyramidaux (FPN)<\/strong>. Voici l\u2019architecture d\u2019un mod\u00e8le tr\u00e8s performant appel\u00e9 TextSnake.<\/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-7a8dda2 elementor-widget elementor-widget-image\" data-id=\"7a8dda2\" 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.google.com\/url?q=https:\/\/github.com\/princewang1994\/TextSnake.pytorch&#038;sa=D&#038;source=docs&#038;ust=1663177989032936&#038;usg=AOvVaw0fOc0VjJpdcTIgPDfPbGPe\">\n\t\t\t\t\t\t\t<img decoding=\"async\" width=\"800\" height=\"486\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2022\/10\/TextSnake-OCR-1-1024x622.jpg\" class=\"attachment-large size-large wp-image-178391\" alt=\"\" loading=\"lazy\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2022\/10\/TextSnake-OCR-1-1024x622.jpg 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2022\/10\/TextSnake-OCR-1-300x182.jpg 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2022\/10\/TextSnake-OCR-1-768x466.jpg 768w, https:\/\/datascientest.com\/wp-content\/uploads\/2022\/10\/TextSnake-OCR-1-1536x933.jpg 1536w, https:\/\/datascientest.com\/wp-content\/uploads\/2022\/10\/TextSnake-OCR-1.jpg 1734w\" 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<figcaption class=\"widget-image-caption wp-caption-text\">Architecture et exemple de segmentation du mod\u00e8le TextSnake<\/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\t\t<\/div>\n\t\t<div class=\"elementor-element elementor-element-8ce2b09 e-flex e-con-boxed e-con\" data-id=\"8ce2b09\" 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-0adaee4 elementor-widget elementor-widget-heading\" data-id=\"0adaee4\" 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\">4. Reconnaissance de texte \u00e0 l\u2019aide du Deep Learning<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-973ab9a elementor-widget elementor-widget-text-editor\" data-id=\"973ab9a\" 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 venons de voir la premi\u00e8re \u00e9tape d\u2019un mod\u00e8le d\u2019oc\u00e9risation qui est la d\u00e9tection des zones du document comportant du texte. Une fois cette \u00e9tape r\u00e9alis\u00e9e, il faut reconna\u00eetre les caract\u00e8res qui composent ces zones pour en d\u00e9duire des mots et ainsi traduire le document manuscrit en un format texte exploitable. <strong>Les mod\u00e8les de Machine Learning les plus utilis\u00e9s pour r\u00e9soudre les probl\u00e8mes de reconnaissance texte sont les r\u00e9seaux de neurones r\u00e9currents<\/strong>. Leurs avantages sont que les couches cach\u00e9es qui les composent partagent une bande m\u00e9moire en commun ce qui a pour cons\u00e9quence d\u2019influencer la pr\u00e9diction en fonction des pr\u00e9c\u00e9dentes pr\u00e9dictions. Si vous voulez en savoir plus sur leur fonctionnement, <a href=\"https:\/\/datascientest.com\/recurrent-neural-network\">un article complet sur les r\u00e9seaux de neurones r\u00e9currents<\/a> est disponible sur notre blog. Les couches les plus utilis\u00e9es pour les mod\u00e8les de r\u00e9seaux de neurones r\u00e9currents sont les <strong>LSTMs<\/strong> et les <strong>GRUs<\/strong>. Nous ajoutons \u00e0 ces couches r\u00e9currentes des couches convolutionnelles. La convolution permet d\u2019extraire des caract\u00e9ristiques pertinentes et locales des images comportant du texte et les couches r\u00e9currentes permettent d\u2019attribuer des labels correspondant aux caract\u00e8res pr\u00e9sents sur l\u2019image. Pour chaque caract\u00e9ristique cr\u00e9\u00e9e par la convolution va \u00eatre associ\u00e9e un vecteur probabilit\u00e9 d\u2019appartenance \u00e0 chaque label.<\/p><p>On peut r\u00e9sumer la reconnaissance de texte gr\u00e2ce au sch\u00e9ma suivant :<\/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-936ae74 elementor-widget elementor-widget-image\" data-id=\"936ae74\" 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=\"1197\" height=\"1053\" src=\"https:\/\/datascientest.com\/wp-content\/uploads\/2022\/10\/reconnaissance-text-dl-OCR-1.jpg\" class=\"attachment-full size-full wp-image-178392\" alt=\"reconnaissance de texte OCR\" loading=\"lazy\" srcset=\"https:\/\/datascientest.com\/wp-content\/uploads\/2022\/10\/reconnaissance-text-dl-OCR-1.jpg 1197w, https:\/\/datascientest.com\/wp-content\/uploads\/2022\/10\/reconnaissance-text-dl-OCR-1-300x264.jpg 300w, https:\/\/datascientest.com\/wp-content\/uploads\/2022\/10\/reconnaissance-text-dl-OCR-1-1024x901.jpg 1024w, https:\/\/datascientest.com\/wp-content\/uploads\/2022\/10\/reconnaissance-text-dl-OCR-1-768x676.jpg 768w\" sizes=\"(max-width: 1197px) 100vw, 1197px\" \/>\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-fae115f elementor-widget elementor-widget-text-editor\" data-id=\"fae115f\" 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>Il ne faut pas oublier qu\u2019il doit y avoir une partie d\u00e9tection de texte en amont pour que la partie reconnaissance fonctionne correctement.<\/p><p>Sur cet exemple, on voit que l\u2019image donn\u00e9e en entr\u00e9e repr\u00e9sente le texte <strong>\u201ca b\u201d<\/strong>. <strong>La partie<\/strong><br \/><strong>convolution va d\u00e9composer l\u2019image en feature map, aussi appel\u00e9e carte des caract\u00e9ristiques<\/strong>. On voit qu\u2019ici l\u2019image a \u00e9t\u00e9 d\u00e9coup\u00e9e en 5 s\u00e9quences. Les couches r\u00e9currentes vont ensuite associer une probabilit\u00e9 de classe \u00e0 chaque s\u00e9quence. Il y a autant de classes que de caract\u00e8res possibles (lettres de l\u2019alphabet, caract\u00e8res sp\u00e9ciaux, \u2026). Pour chaque s\u00e9quence, on choisit la classe qui maximise la probabilit\u00e9, on obtient une suite de caract\u00e8res et on peut ensuite en d\u00e9duire notre texte final.<\/p><p>Cependant, ce mod\u00e8le ne permet pas de v\u00e9rifier que le mot ou la phrase retourn\u00e9e en sortie soit <strong>sans faute d&rsquo;orthographe<\/strong> car on pr\u00e9dit caract\u00e8re par caract\u00e8re sans se soucier de cela. Il existe d\u2019autres mod\u00e8les qui mettent en place un <strong>syst\u00e8me de correction orthographique<\/strong>. Cette approche s\u2019appelle un mod\u00e8le de langage, et peut am\u00e9liorer la pr\u00e9cision de notre mod\u00e8le.<\/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-3208216 e-flex e-con-boxed e-con\" data-id=\"3208216\" 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-412ff02 elementor-widget elementor-widget-heading\" data-id=\"412ff02\" 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\">5. Conclusion<\/h3>\t\t<\/div>\n\t\t\t\t<\/div>\n\t\t\t\t<div class=\"elementor-element elementor-element-d06ab97 elementor-widget elementor-widget-text-editor\" data-id=\"d06ab97\" 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 venons de voir un <strong>mod\u00e8le de Deep Learning<\/strong> permettant de r\u00e9soudre le probl\u00e8me de l\u2019oc\u00e9risation. Le Deep Learning a vraiment r\u00e9volutionn\u00e9 ce domaine car il offre des <strong>pr\u00e9dictions<\/strong> tr\u00e8s pr\u00e9cises contrairement aux mod\u00e8les qui avaient \u00e9t\u00e9 mis au point auparavant. On peut remarquer que le mod\u00e8le utilise des techniques normalement utilis\u00e9es dans d\u2019autres domaines comme la <strong>d\u00e9tection d\u2019objets dans une image<\/strong> ou encore le <a href=\"https:\/\/datascientest.com\/introduction-au-nlp-natural-language-processing\">Natural Language Processing<\/a> et que ces techniques fonctionnent tr\u00e8s bien pour l\u2019oc\u00e9risation.<\/p><p>Si vous avez aim\u00e9 cet article et que vous voulez d\u00e9couvrir d\u2019<a href=\"https:\/\/datascientest.com\/deep-learning-avec-python-decouvrez-les-fondamentaux\">autres m\u00e9thodes de Deep Learning<\/a>, je vous invite \u00e0 rejoindre notre cursus expert en Deep Learning.<\/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-bdabcf9 elementor-align-center button-principale elementor-widget elementor-widget-button\" data-id=\"bdabcf9\" 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-deep-learning\">\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 la formation Deep Learning<\/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\t\t<\/div>\n\t\t\t\t\t\t\t<\/div>\n\t\t","protected":false},"excerpt":{"rendered":"<p>La reconnaissance optique de caract\u00e8res (ROC ou Optical Character Recognition en anglais), appel\u00e9e \u00e9galement oc\u00e9risation, regroupe l\u2019ensemble des m\u00e9thodes qui permettent de g\u00e9n\u00e9rer des fichiers textes \u00e0 partir d\u2019images contenant du texte manuscrit. Avec l\u2019av\u00e8nement du num\u00e9rique et de l\u2019automatisation, l&rsquo;oc\u00e9risation est devenue un outil incontournable car les images comportant du texte ne sont pas [&hellip;]<\/p>\n","protected":false},"author":77,"featured_media":177535,"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>Apprendre \u00e0 l\u2019ordinateur \u00e0 lire avec l\u2019Optical Character Recognition<\/title>\n<meta name=\"description\" content=\"La reconnaissance optique de caract\u00e8res (ROC ou Optical Character Recognition en anglais), appel\u00e9e \u00e9galement oc\u00e9risation, regroupe l\u2019ensemble des m\u00e9thodes qui permettent de g\u00e9n\u00e9rer des fichiers textes \u00e0 partir d\u2019images contenant du texte manuscrit.\" \/>\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\/optical-character-recognition\" \/>\n<meta property=\"og:locale\" content=\"fr_FR\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"Apprendre \u00e0 l\u2019ordinateur \u00e0 lire avec l\u2019Optical Character Recognition\" \/>\n<meta property=\"og:url\" content=\"https:\/\/datascientest.com\/optical-character-recognition\" \/>\n<meta property=\"og:site_name\" content=\"Formation Data Science | DataScientest.com\" \/>\n<meta property=\"article:published_time\" content=\"2022-10-14T09:46:44+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2023-11-09T13:56:11+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/datascientest.com\/wp-content\/uploads\/2022\/09\/Illu_BLOG_NEW_OCR.png\" \/>\n\t<meta property=\"og:image:width\" content=\"2134\" \/>\n\t<meta property=\"og:image:height\" content=\"1171\" \/>\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=\"7 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/datascientest.com\/optical-character-recognition\",\"url\":\"https:\/\/datascientest.com\/optical-character-recognition\",\"name\":\"Apprendre \u00e0 l\u2019ordinateur \u00e0 lire avec l\u2019Optical Character Recognition\",\"isPartOf\":{\"@id\":\"https:\/\/datascientest.com\/#website\"},\"datePublished\":\"2022-10-14T09:46:44+00:00\",\"dateModified\":\"2023-11-09T13:56:11+00:00\",\"author\":{\"@id\":\"https:\/\/datascientest.com\/#\/schema\/person\/6eca16f63daaafcbc96e15549b9d4ee1\"},\"description\":\"La reconnaissance optique de caract\u00e8res (ROC ou Optical Character Recognition en anglais), appel\u00e9e \u00e9galement oc\u00e9risation, regroupe l\u2019ensemble des m\u00e9thodes qui permettent de g\u00e9n\u00e9rer des fichiers textes \u00e0 partir d\u2019images contenant du texte manuscrit.\",\"inLanguage\":\"fr-FR\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/datascientest.com\/optical-character-recognition\"]}]},{\"@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":"Apprendre \u00e0 l\u2019ordinateur \u00e0 lire avec l\u2019Optical Character Recognition","description":"La reconnaissance optique de caract\u00e8res (ROC ou Optical Character Recognition en anglais), appel\u00e9e \u00e9galement oc\u00e9risation, regroupe l\u2019ensemble des m\u00e9thodes qui permettent de g\u00e9n\u00e9rer des fichiers textes \u00e0 partir d\u2019images contenant du texte manuscrit.","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\/optical-character-recognition","og_locale":"fr_FR","og_type":"article","og_title":"Apprendre \u00e0 l\u2019ordinateur \u00e0 lire avec l\u2019Optical Character Recognition","og_url":"https:\/\/datascientest.com\/optical-character-recognition","og_site_name":"Formation Data Science | DataScientest.com","article_published_time":"2022-10-14T09:46:44+00:00","article_modified_time":"2023-11-09T13:56:11+00:00","og_image":[{"width":2134,"height":1171,"url":"https:\/\/datascientest.com\/wp-content\/uploads\/2022\/09\/Illu_BLOG_NEW_OCR.png","type":"image\/png"}],"author":"Raphael Kassel","twitter_card":"summary_large_image","twitter_misc":{"\u00c9crit par":"Raphael Kassel","Dur\u00e9e de lecture estim\u00e9e":"7 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/datascientest.com\/optical-character-recognition","url":"https:\/\/datascientest.com\/optical-character-recognition","name":"Apprendre \u00e0 l\u2019ordinateur \u00e0 lire avec l\u2019Optical Character Recognition","isPartOf":{"@id":"https:\/\/datascientest.com\/#website"},"datePublished":"2022-10-14T09:46:44+00:00","dateModified":"2023-11-09T13:56:11+00:00","author":{"@id":"https:\/\/datascientest.com\/#\/schema\/person\/6eca16f63daaafcbc96e15549b9d4ee1"},"description":"La reconnaissance optique de caract\u00e8res (ROC ou Optical Character Recognition en anglais), appel\u00e9e \u00e9galement oc\u00e9risation, regroupe l\u2019ensemble des m\u00e9thodes qui permettent de g\u00e9n\u00e9rer des fichiers textes \u00e0 partir d\u2019images contenant du texte manuscrit.","inLanguage":"fr-FR","potentialAction":[{"@type":"ReadAction","target":["https:\/\/datascientest.com\/optical-character-recognition"]}]},{"@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\/176176"}],"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=176176"}],"version-history":[{"count":0,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/posts\/176176\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/media\/177535"}],"wp:attachment":[{"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/media?parent=176176"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/categories?post=176176"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/datascientest.com\/wp-json\/wp\/v2\/tags?post=176176"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}