{"id":443338,"date":"2024-02-07T20:02:29","date_gmt":"2024-02-07T19:02:29","guid":{"rendered":"https:\/\/www.eenewseurope.com\/?p=443338"},"modified":"2024-02-07T20:02:29","modified_gmt":"2024-02-07T19:02:29","slug":"lepfl-met-lia-au-service-de-la-compression-dimages","status":"publish","type":"post","link":"https:\/\/www.ecinews.fr\/fr\/lepfl-met-lia-au-service-de-la-compression-dimages\/","title":{"rendered":"L’EPFL met l’IA au service de la compression d’images"},"content":{"rendered":"

Des chercheurs suisses ont utilis\u00e9 un jumeau num\u00e9rique \u00e0 apprentissage automatique pour compresser des donn\u00e9es d’image avec une plus grande pr\u00e9cision que les m\u00e9thodes de calcul sans apprentissage.<\/p>\n

La technique de compression d’images AI d\u00e9velopp\u00e9e \u00e0 l’EPFL a des applications pour les implants r\u00e9tiniens et d’autres produits en \u00e9lectronique m\u00e9dicale.<\/p>\n

L’un des principaux d\u00e9fis \u00e0 relever pour d\u00e9velopper de meilleures proth\u00e8ses neurales est le codage sensoriel : il s’agit de transformer les informations capt\u00e9es dans l’environnement par des capteurs en signaux neuronaux susceptibles d’\u00eatre interpr\u00e9t\u00e9s par le syst\u00e8me nerveux. Mais comme le nombre d’\u00e9lectrodes d’une proth\u00e8se est limit\u00e9, cet apport environnemental doit \u00eatre r\u00e9duit d’une mani\u00e8re ou d’une autre, tout en pr\u00e9servant la qualit\u00e9 des donn\u00e9es transmises au cerveau.<\/p>\n

Demetri Psaltis du Laboratoire d’optique de l’EPFL et Christophe Moser du Laboratoire des dispositifs photoniques appliqu\u00e9s ont collabor\u00e9 avec Diego Ghezzi de l’H\u00f4pital ophtalmique Jules-Gonin – Fondation Asile des Aveugles (pr\u00e9c\u00e9demment titulaire de la chaire Medtronic de neuro-ing\u00e9nierie \u00e0 l’EPFL) pour appliquer l’apprentissage automatique au probl\u00e8me de la compression des donn\u00e9es d’images \u00e0 dimensions multiples, telles que la couleur et le contraste.<\/p>\n

Dans ce cas, l’objectif de la compression \u00e9tait le sous-\u00e9chantillonnage, c’est-\u00e0-dire la r\u00e9duction du nombre de pixels d’une image \u00e0 transmettre par l’interm\u00e9diaire d’une proth\u00e8se r\u00e9tinienne.<\/p>\n

\u00ab\u00a0Le sous-\u00e9chantillonnage pour les implants r\u00e9tiniens se fait actuellement par le biais d’une moyenne de pixels, ce qui est essentiellement ce que font les logiciels graphiques lorsque vous voulez r\u00e9duire la taille d’un fichier. Mais en fin de compte, il s’agit d’un processus math\u00e9matique ; il n’y a pas d’apprentissage en jeu\u00a0\u00bb, explique M. Ghezzi.<\/p>\n

\u00ab\u00a0Nous avons constat\u00e9 que si nous appliquions une approche bas\u00e9e sur l’apprentissage, nous obtenions de meilleurs r\u00e9sultats en termes d’encodage sensoriel optimis\u00e9. Mais ce qui est plus surprenant, c’est que lorsque nous avons utilis\u00e9 un r\u00e9seau neuronal non contraint, il a appris \u00e0 imiter de lui-m\u00eame certains aspects du traitement r\u00e9tinien\u00a0\u00bb.<\/p>\n

La technique de compression d’images de l’IA ou \u00ab\u00a0mod\u00e8le associatif\u00a0\u00bb permet de trouver un point id\u00e9al pour le contraste de l’image.<\/h4>\n

Dans le cadre de l’IA du mod\u00e8le associatif, deux r\u00e9seaux neuronaux fonctionnent de mani\u00e8re compl\u00e9mentaire. La partie mod\u00e8le, ou mod\u00e8le avant, agit comme un jumeau num\u00e9rique de la r\u00e9tine : elle est d’abord entra\u00een\u00e9e \u00e0 recevoir une image haute r\u00e9solution et \u00e0 produire un code neuronal binaire aussi proche que possible du code neuronal g\u00e9n\u00e9r\u00e9 par une r\u00e9tine biologique.<\/p>\n

Le mod\u00e8le associatif est ensuite entra\u00een\u00e9 \u00e0 r\u00e9duire l’\u00e9chantillonnage d’une image \u00e0 haute r\u00e9solution \u00e0 l’aide du mod\u00e8le direct qui est aussi proche que possible de celui produit par la r\u00e9tine biologique en r\u00e9ponse \u00e0 l’image d’origine.<\/p>\n

\u00ab\u00a0La prochaine \u00e9tape \u00e9vidente est de voir comment nous pouvons compresser une image de mani\u00e8re plus large, au-del\u00e0 de la r\u00e9duction des pixels, afin que le cadre puisse jouer avec plusieurs dimensions visuelles en m\u00eame temps. Une autre possibilit\u00e9 consiste \u00e0 transposer ce mod\u00e8le r\u00e9tinien \u00e0 des sorties provenant d’autres r\u00e9gions du cerveau. Il pourrait m\u00eame \u00eatre reli\u00e9 \u00e0 d’autres dispositifs, comme des proth\u00e8ses auditives ou des proth\u00e8ses de membres\u00a0\u00bb, a d\u00e9clar\u00e9 M. Ghezzi.<\/p>\n

Nature communications<\/a><\/p>\n

EPFL<\/a><\/p>\n

Suivre ECInews sur Google news<\/a><\/p>\n

 <\/p>\n

 <\/p>\n","protected":false},"excerpt":{"rendered":"

Des chercheurs suisses ont utilis\u00e9 un jumeau num\u00e9rique \u00e0 apprentissage automatique pour compresser des donn\u00e9es d’image avec une plus grande pr\u00e9cision que les m\u00e9thodes de calcul sans apprentissage. La technique de compression d’images AI d\u00e9velopp\u00e9e \u00e0 l’EPFL a des applications pour les implants r\u00e9tiniens et d’autres produits en \u00e9lectronique m\u00e9dicale. L’un des principaux d\u00e9fis \u00e0 […]<\/p>\n","protected":false},"author":11,"featured_media":443298,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_acf_changed":false,"footnotes":""},"categories":[883],"tags":[1652,5447,2198],"domains":[47],"ppma_author":[1143,3640],"class_list":["post-443338","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-technologies","tag-electronique-medicale","tag-epfl-fr","tag-ia","domains-electronique-eci"],"acf":[],"yoast_head":"L'EPFL met l'IA au service de la compression d'images ...<\/title>\n<meta name=\"description\" content=\"Compresser des images avec une plus grande pr\u00e9cision que les m\u00e9thodes de calcul sans apprentissage pour les conceptions m\u00e9dicales.\" \/>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.ecinews.fr\/fr\/wp-json\/wp\/v2\/posts\/443338\/\" \/>\n<meta property=\"og:locale\" content=\"fr_FR\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"L'EPFL met l'IA au service de la compression d'images\" \/>\n<meta property=\"og:description\" content=\"Compresser des images avec une plus grande pr\u00e9cision que les m\u00e9thodes de calcul sans apprentissage pour les conceptions m\u00e9dicales.\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.ecinews.fr\/fr\/wp-json\/wp\/v2\/posts\/443338\/\" \/>\n<meta property=\"og:site_name\" content=\"EENewsEurope\" \/>\n<meta property=\"article:published_time\" content=\"2024-02-07T19:02:29+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/www.ecinews.fr\/wp-content\/uploads\/2024\/02\/EPFL-image-compression-scaled.jpg\" \/>\n\t<meta property=\"og:image:width\" content=\"1080\" \/>\n\t<meta property=\"og:image:height\" content=\"606\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\n<meta name=\"author\" content=\"Daniel Cardon, Nick Flaherty\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Daniel Cardon\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"3 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"Article\",\"@id\":\"https:\/\/www.ecinews.fr\/fr\/lepfl-met-lia-au-service-de-la-compression-dimages\/#article\",\"isPartOf\":{\"@id\":\"https:\/\/www.ecinews.fr\/fr\/lepfl-met-lia-au-service-de-la-compression-dimages\/\"},\"author\":{\"name\":\"Daniel Cardon\",\"@id\":\"https:\/\/www.ecinews.fr\/fr\/#\/schema\/person\/93c44cde463762f40a8236eaa44c1c17\"},\"headline\":\"L’EPFL met l’IA au service de la compression d’images\",\"datePublished\":\"2024-02-07T19:02:29+00:00\",\"dateModified\":\"2024-02-07T19:02:29+00:00\",\"mainEntityOfPage\":{\"@id\":\"https:\/\/www.ecinews.fr\/fr\/lepfl-met-lia-au-service-de-la-compression-dimages\/\"},\"wordCount\":626,\"publisher\":{\"@id\":\"https:\/\/www.ecinews.fr\/fr\/#organization\"},\"keywords\":[\"Electronique m\u00e9dicale\",\"epfl\",\"IA\"],\"articleSection\":[\"Technologies\"],\"inLanguage\":\"fr-FR\"},{\"@type\":\"WebPage\",\"@id\":\"https:\/\/www.ecinews.fr\/fr\/lepfl-met-lia-au-service-de-la-compression-dimages\/\",\"url\":\"https:\/\/www.ecinews.fr\/fr\/lepfl-met-lia-au-service-de-la-compression-dimages\/\",\"name\":\"L'EPFL met l'IA au service de la compression d'images -\",\"isPartOf\":{\"@id\":\"https:\/\/www.ecinews.fr\/fr\/#website\"},\"datePublished\":\"2024-02-07T19:02:29+00:00\",\"dateModified\":\"2024-02-07T19:02:29+00:00\",\"description\":\"Compresser des images avec une plus grande pr\u00e9cision que les m\u00e9thodes de calcul sans apprentissage pour les conceptions m\u00e9dicales.\",\"breadcrumb\":{\"@id\":\"https:\/\/www.ecinews.fr\/fr\/lepfl-met-lia-au-service-de-la-compression-dimages\/#breadcrumb\"},\"inLanguage\":\"fr-FR\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/www.ecinews.fr\/fr\/lepfl-met-lia-au-service-de-la-compression-dimages\/\"]}]},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/www.ecinews.fr\/fr\/lepfl-met-lia-au-service-de-la-compression-dimages\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Home\",\"item\":\"https:\/\/www.ecinews.fr\/fr\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"L’EPFL met l’IA au service de la compression d’images\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/www.ecinews.fr\/fr\/#website\",\"url\":\"https:\/\/www.ecinews.fr\/fr\/\",\"name\":\"EENewsEurope\",\"description\":\"Just another WordPress site\",\"publisher\":{\"@id\":\"https:\/\/www.ecinews.fr\/fr\/#organization\"},\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/www.ecinews.fr\/fr\/?s={search_term_string}\"},\"query-input\":\"required name=search_term_string\"}],\"inLanguage\":\"fr-FR\"},{\"@type\":\"Organization\",\"@id\":\"https:\/\/www.ecinews.fr\/fr\/#organization\",\"name\":\"EENewsEurope\",\"url\":\"https:\/\/www.ecinews.fr\/fr\/\",\"logo\":{\"@type\":\"ImageObject\",\"inLanguage\":\"fr-FR\",\"@id\":\"https:\/\/www.ecinews.fr\/fr\/#\/schema\/logo\/image\/\",\"url\":\"https:\/\/www.ecinews.fr\/wp-content\/uploads\/2022\/02\/logo-1.jpg\",\"contentUrl\":\"https:\/\/www.ecinews.fr\/wp-content\/uploads\/2022\/02\/logo-1.jpg\",\"width\":283,\"height\":113,\"caption\":\"EENewsEurope\"},\"image\":{\"@id\":\"https:\/\/www.ecinews.fr\/fr\/#\/schema\/logo\/image\/\"}},{\"@type\":\"Person\",\"@id\":\"https:\/\/www.ecinews.fr\/fr\/#\/schema\/person\/93c44cde463762f40a8236eaa44c1c17\",\"name\":\"Daniel Cardon\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"fr-FR\",\"@id\":\"https:\/\/www.ecinews.fr\/fr\/#\/schema\/person\/image\/2b243f6bcc1cff7d86aadfb2cd0bd870\",\"url\":\"https:\/\/secure.gravatar.com\/avatar\/de9edc136dfc061ad4778e04635baefa?s=96&d=mm&r=g\",\"contentUrl\":\"https:\/\/secure.gravatar.com\/avatar\/de9edc136dfc061ad4778e04635baefa?s=96&d=mm&r=g\",\"caption\":\"Daniel Cardon\"}}]}<\/script>","yoast_head_json":{"title":"L'EPFL met l'IA au service de la compression d'images ...","description":"Compresser des images avec une plus grande pr\u00e9cision que les m\u00e9thodes de calcul sans apprentissage pour les conceptions m\u00e9dicales.","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:\/\/www.ecinews.fr\/fr\/wp-json\/wp\/v2\/posts\/443338\/","og_locale":"fr_FR","og_type":"article","og_title":"L'EPFL met l'IA au service de la compression d'images","og_description":"Compresser des images avec une plus grande pr\u00e9cision que les m\u00e9thodes de calcul sans apprentissage pour les conceptions m\u00e9dicales.","og_url":"https:\/\/www.ecinews.fr\/fr\/wp-json\/wp\/v2\/posts\/443338\/","og_site_name":"EENewsEurope","article_published_time":"2024-02-07T19:02:29+00:00","og_image":[{"width":1080,"height":606,"url":"https:\/\/www.ecinews.fr\/wp-content\/uploads\/2024\/02\/EPFL-image-compression-scaled.jpg","type":"image\/jpeg"}],"author":"Daniel Cardon, Nick Flaherty","twitter_card":"summary_large_image","twitter_misc":{"Written by":"Daniel Cardon","Est. reading time":"3 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"Article","@id":"https:\/\/www.ecinews.fr\/fr\/lepfl-met-lia-au-service-de-la-compression-dimages\/#article","isPartOf":{"@id":"https:\/\/www.ecinews.fr\/fr\/lepfl-met-lia-au-service-de-la-compression-dimages\/"},"author":{"name":"Daniel Cardon","@id":"https:\/\/www.ecinews.fr\/fr\/#\/schema\/person\/93c44cde463762f40a8236eaa44c1c17"},"headline":"L’EPFL met l’IA au service de la compression d’images","datePublished":"2024-02-07T19:02:29+00:00","dateModified":"2024-02-07T19:02:29+00:00","mainEntityOfPage":{"@id":"https:\/\/www.ecinews.fr\/fr\/lepfl-met-lia-au-service-de-la-compression-dimages\/"},"wordCount":626,"publisher":{"@id":"https:\/\/www.ecinews.fr\/fr\/#organization"},"keywords":["Electronique m\u00e9dicale","epfl","IA"],"articleSection":["Technologies"],"inLanguage":"fr-FR"},{"@type":"WebPage","@id":"https:\/\/www.ecinews.fr\/fr\/lepfl-met-lia-au-service-de-la-compression-dimages\/","url":"https:\/\/www.ecinews.fr\/fr\/lepfl-met-lia-au-service-de-la-compression-dimages\/","name":"L'EPFL met l'IA au service de la compression d'images -","isPartOf":{"@id":"https:\/\/www.ecinews.fr\/fr\/#website"},"datePublished":"2024-02-07T19:02:29+00:00","dateModified":"2024-02-07T19:02:29+00:00","description":"Compresser des images avec une plus grande pr\u00e9cision que les m\u00e9thodes de calcul sans apprentissage pour les conceptions m\u00e9dicales.","breadcrumb":{"@id":"https:\/\/www.ecinews.fr\/fr\/lepfl-met-lia-au-service-de-la-compression-dimages\/#breadcrumb"},"inLanguage":"fr-FR","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.ecinews.fr\/fr\/lepfl-met-lia-au-service-de-la-compression-dimages\/"]}]},{"@type":"BreadcrumbList","@id":"https:\/\/www.ecinews.fr\/fr\/lepfl-met-lia-au-service-de-la-compression-dimages\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https:\/\/www.ecinews.fr\/fr\/"},{"@type":"ListItem","position":2,"name":"L’EPFL met l’IA au service de la compression d’images"}]},{"@type":"WebSite","@id":"https:\/\/www.ecinews.fr\/fr\/#website","url":"https:\/\/www.ecinews.fr\/fr\/","name":"EENewsEurope","description":"Just another WordPress site","publisher":{"@id":"https:\/\/www.ecinews.fr\/fr\/#organization"},"potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.ecinews.fr\/fr\/?s={search_term_string}"},"query-input":"required name=search_term_string"}],"inLanguage":"fr-FR"},{"@type":"Organization","@id":"https:\/\/www.ecinews.fr\/fr\/#organization","name":"EENewsEurope","url":"https:\/\/www.ecinews.fr\/fr\/","logo":{"@type":"ImageObject","inLanguage":"fr-FR","@id":"https:\/\/www.ecinews.fr\/fr\/#\/schema\/logo\/image\/","url":"https:\/\/www.ecinews.fr\/wp-content\/uploads\/2022\/02\/logo-1.jpg","contentUrl":"https:\/\/www.ecinews.fr\/wp-content\/uploads\/2022\/02\/logo-1.jpg","width":283,"height":113,"caption":"EENewsEurope"},"image":{"@id":"https:\/\/www.ecinews.fr\/fr\/#\/schema\/logo\/image\/"}},{"@type":"Person","@id":"https:\/\/www.ecinews.fr\/fr\/#\/schema\/person\/93c44cde463762f40a8236eaa44c1c17","name":"Daniel Cardon","image":{"@type":"ImageObject","inLanguage":"fr-FR","@id":"https:\/\/www.ecinews.fr\/fr\/#\/schema\/person\/image\/2b243f6bcc1cff7d86aadfb2cd0bd870","url":"https:\/\/secure.gravatar.com\/avatar\/de9edc136dfc061ad4778e04635baefa?s=96&d=mm&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/de9edc136dfc061ad4778e04635baefa?s=96&d=mm&r=g","caption":"Daniel Cardon"}}]}},"authors":[{"term_id":1143,"user_id":11,"is_guest":0,"slug":"danielcardon","display_name":"Daniel Cardon","avatar_url":"https:\/\/secure.gravatar.com\/avatar\/de9edc136dfc061ad4778e04635baefa?s=96&d=mm&r=g","0":null,"1":"","2":"","3":"","4":"","5":"","6":"","7":"","8":""},{"term_id":3640,"user_id":0,"is_guest":1,"slug":"nick-flaherty","display_name":"Nick Flaherty","avatar_url":"https:\/\/secure.gravatar.com\/avatar\/?s=96&d=mm&r=g","0":null,"1":"","2":"","3":"","4":"","5":"","6":"","7":"","8":""}],"_links":{"self":[{"href":"https:\/\/www.ecinews.fr\/fr\/wp-json\/wp\/v2\/posts\/443338"}],"collection":[{"href":"https:\/\/www.ecinews.fr\/fr\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.ecinews.fr\/fr\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.ecinews.fr\/fr\/wp-json\/wp\/v2\/users\/11"}],"replies":[{"embeddable":true,"href":"https:\/\/www.ecinews.fr\/fr\/wp-json\/wp\/v2\/comments?post=443338"}],"version-history":[{"count":0,"href":"https:\/\/www.ecinews.fr\/fr\/wp-json\/wp\/v2\/posts\/443338\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.ecinews.fr\/fr\/wp-json\/wp\/v2\/media\/443298"}],"wp:attachment":[{"href":"https:\/\/www.ecinews.fr\/fr\/wp-json\/wp\/v2\/media?parent=443338"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.ecinews.fr\/fr\/wp-json\/wp\/v2\/categories?post=443338"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.ecinews.fr\/fr\/wp-json\/wp\/v2\/tags?post=443338"},{"taxonomy":"domains","embeddable":true,"href":"https:\/\/www.ecinews.fr\/fr\/wp-json\/wp\/v2\/domains?post=443338"},{"taxonomy":"author","embeddable":true,"href":"https:\/\/www.ecinews.fr\/fr\/wp-json\/wp\/v2\/ppma_author?post=443338"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}