{"id":906,"date":"2025-07-06T05:28:51","date_gmt":"2025-07-06T05:28:51","guid":{"rendered":"https:\/\/janusai.pro\/?p=906"},"modified":"2025-07-06T05:28:52","modified_gmt":"2025-07-06T05:28:52","slug":"janus-4o-sharegpt-4o-image","status":"publish","type":"post","link":"https:\/\/janusai.pro\/da\/janus-4o-sharegpt-4o-image\/","title":{"rendered":"Den nye stjerne inden for multimodal billedgenerering: Janus-4o? DelGPT-4o-Image s\u00e6tter en ny standard for datas\u00e6t og tilpasser billedgenerering til GPT-4o."},"content":{"rendered":"<div style=\"margin-top: 0px; margin-bottom: 0px;\" class=\"sharethis-inline-share-buttons\" ><\/div>\n<p><a href=\"https:\/\/sharegpt4o.github.io\/\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">DelGPT-4o-billede<\/a> er et storstilet datas\u00e6t til billedgenerering af h\u00f8j kvalitet, hvor alle billeder genereres ved hj\u00e6lp af GPT-4os billedgenereringsfunktioner.<\/p>\n\n\n\n<p>Dette datas\u00e6t har til form\u00e5l at kombinere fordelene ved open source multimodale modeller med GPT-4os styrker inden for skabelse af visuelt indhold. <\/p>\n\n\n\n<p>Den indeholder 45.000 tekst-til-billede- og 46.000 billede-til-tekst-eksempler, hvilket g\u00f8r den til en praktisk ressource til forbedring af multimodale modeller i forbindelse med billedgenerering og redigeringsopgaver.<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img loading=\"lazy\" decoding=\"async\" width=\"998\" height=\"700\" src=\"https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/f48c8349-9310-48a1-9276-d7614aa958d9.png\" alt=\"\" class=\"wp-image-911\" srcset=\"https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/f48c8349-9310-48a1-9276-d7614aa958d9.png 998w, https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/f48c8349-9310-48a1-9276-d7614aa958d9-300x210.png 300w, https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/f48c8349-9310-48a1-9276-d7614aa958d9-768x539.png 768w, https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/f48c8349-9310-48a1-9276-d7614aa958d9-18x12.png 18w\" sizes=\"auto, (max-width: 998px) 100vw, 998px\" \/><\/figure>\n\n\n\n<p>Janus-4o er en multimodal LLM, der er i stand til at generere tekst-til-billede og tekst+billede-til-billede. Den er baseret p\u00e5 Janus-Pro og finjusteret ved hj\u00e6lp af ShareGPT-4o-Image-datas\u00e6ttet. Sammenlignet med Janus-Pro introducerer Janus-4o tekst+billede-til-billede-genereringsfunktioner og opn\u00e5r betydelige forbedringer i tekst-til-billede-generering.<\/p>\n\n\n\n<div id=\"ez-toc-container\" class=\"ez-toc-v2_0_82_2 counter-hierarchy ez-toc-counter ez-toc-grey ez-toc-container-direction\">\n<div class=\"ez-toc-title-container\">\n<p class=\"ez-toc-title\" style=\"cursor:inherit\">Indholdsfortegnelse<\/p>\n<span class=\"ez-toc-title-toggle\"><a href=\"#\" class=\"ez-toc-pull-right ez-toc-btn ez-toc-btn-xs ez-toc-btn-default ez-toc-toggle\" aria-label=\"Skift til indholdsfortegnelse\"><span class=\"ez-toc-js-icon-con\"><span class=\"\"><span class=\"eztoc-hide\" style=\"display:none;\">Toggle<\/span><span class=\"ez-toc-icon-toggle-span\"><svg style=\"fill: #999;color:#999\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" class=\"list-377408\" width=\"20px\" height=\"20px\" viewbox=\"0 0 24 24\" fill=\"none\"><path d=\"M6 6H4v2h2V6zm14 0H8v2h12V6zM4 11h2v2H4v-2zm16 0H8v2h12v-2zM4 16h2v2H4v-2zm16 0H8v2h12v-2z\" fill=\"currentColor\"><\/path><\/svg><svg style=\"fill: #999;color:#999\" class=\"arrow-unsorted-368013\" xmlns=\"http:\/\/www.w3.org\/2000\/svg\" width=\"10px\" height=\"10px\" viewbox=\"0 0 24 24\" version=\"1.2\" baseprofile=\"tiny\"><path d=\"M18.2 9.3l-6.2-6.3-6.2 6.3c-.2.2-.3.4-.3.7s.1.5.3.7c.2.2.4.3.7.3h11c.3 0 .5-.1.7-.3.2-.2.3-.5.3-.7s-.1-.5-.3-.7zM5.8 14.7l6.2 6.3 6.2-6.3c.2-.2.3-.5.3-.7s-.1-.5-.3-.7c-.2-.2-.4-.3-.7-.3h-11c-.3 0-.5.1-.7.3-.2.2-.3.5-.3.7s.1.5.3.7z\"\/><\/svg><\/span><\/span><\/span><\/a><\/span><\/div>\n<nav><ul class='ez-toc-list ez-toc-list-level-1' ><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-1\" href=\"https:\/\/janusai.pro\/da\/janus-4o-sharegpt-4o-image\/#Dataset_Overview\" >Oversigt over datas\u00e6t<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-2\" href=\"https:\/\/janusai.pro\/da\/janus-4o-sharegpt-4o-image\/#Related_Links\" >Relaterede links<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-3\" href=\"https:\/\/janusai.pro\/da\/janus-4o-sharegpt-4o-image\/#Paper_Introduction\" >Papirintroduktion<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-4\" href=\"https:\/\/janusai.pro\/da\/janus-4o-sharegpt-4o-image\/#Method_Overview\" >Metodeoversigt<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-5\" href=\"https:\/\/janusai.pro\/da\/janus-4o-sharegpt-4o-image\/#Experimental_Results\" >Eksperimentelle resultater<\/a><\/li><li class='ez-toc-page-1 ez-toc-heading-level-2'><a class=\"ez-toc-link ez-toc-heading-6\" href=\"https:\/\/janusai.pro\/da\/janus-4o-sharegpt-4o-image\/#Conclusions\" >Konklusioner<\/a><\/li><\/ul><\/nav><\/div>\n<h2 class=\"wp-block-heading\"><span class=\"ez-toc-section\" id=\"Dataset_Overview\"><\/span>Oversigt over datas\u00e6t<span class=\"ez-toc-section-end\"><\/span><\/h2>\n\n\n\n<p>ShareGPT-4o-Image-datas\u00e6ttet indeholder 91.000 GPT-4o-billedgenereringseksempler, kategoriseret som f\u00f8lger:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Tekst-til-billede: 45.717<\/li>\n\n\n\n<li>Tekst plus billede til billede: 46.539<\/li>\n<\/ul>\n\n\n\n<h2 class=\"wp-block-heading\"><span class=\"ez-toc-section\" id=\"Related_Links\"><\/span>Relaterede links<span class=\"ez-toc-section-end\"><\/span><\/h2>\n\n\n\n<p>Kode: <a href=\"https:\/\/github.com\/FreedomIntelligence\/ShareGPT-4o-Image\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">github klik her<\/a><\/p>\n\n\n\n<p>Model: <a href=\"https:\/\/huggingface.co\/FreedomIntelligence\/Janus-4o-7B\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">Hent ShareGPT-4o-Image-modellen<\/a><\/p>\n\n\n\n<p>Papir: <a href=\"https:\/\/arxiv.org\/pdf\/2506.18095\" target=\"_blank\" rel=\"noreferrer noopener nofollow\">klik her<\/a><\/p>\n\n\n\n<h2 class=\"wp-block-heading\"><span class=\"ez-toc-section\" id=\"Paper_Introduction\"><\/span>Papirintroduktion<span class=\"ez-toc-section-end\"><\/span><\/h2>\n\n\n\n<p>Nylige fremskridt inden for multimodale genereringsmodeller har \u00e5bnet op for realistisk, instruktionsjusteret billedgenerering. Imidlertid forbliver f\u00f8rende systemer som GPT-4o-Image propriet\u00e6re og utilg\u00e6ngelige.<\/p>\n\n\n\n<p>For at g\u00f8re disse funktioner tilg\u00e6ngelige for offentligheden introducerer artiklen ShareGPT-4o-Image, det f\u00f8rste datas\u00e6t, der indeholder 45.000 tekst-til-billede og 46.000 tekst-plus-billede-til-billede eksempler, alle syntetiseret ved hj\u00e6lp af GPT-4os billedgenereringsfunktioner for at forfine dets avancerede billedgenereringsevner. Ved hj\u00e6lp af dette datas\u00e6t udviklede artiklen Janus-4o, en multimodal stor sprogmodel, der er i stand til tekst-til-billede og tekst-plus-billede-til-billede generering.<\/p>\n\n\n\n<p>Janus-4o forbedrer ikke blot tekst-til-billede-genereringsfunktionerne betydeligt i forhold til sin forg\u00e6nger Janus-Pro, men introducerer ogs\u00e5 tekst-plus-billede-til-billede-genereringsfunktioner. Bem\u00e6rkelsesv\u00e6rdigt opn\u00e5r den imponerende ydeevne i at generere billeder fra tekst og billeder fra bunden ved hj\u00e6lp af kun 91K syntetiske pr\u00f8ver og tr\u00e6net i 6 timer p\u00e5 en 8\u00d7A800 GPU-maskine.<\/p>\n\n\n\n<p>Vi h\u00e5ber, at udgivelsen af ShareGPT-4o-Image og Janus-4o vil fremme \u00e5ben forskning inden for fotorealistisk, instruktionsafstemt billedgenerering.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\"><span class=\"ez-toc-section\" id=\"Method_Overview\"><\/span>Metodeoversigt<span class=\"ez-toc-section-end\"><\/span><\/h2>\n\n\n\n<figure class=\"wp-block-image\"><img loading=\"lazy\" decoding=\"async\" width=\"1028\" height=\"718\" src=\"https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/74bd55e5-5cc6-49e8-be21-cf5c4a66042d.png\" alt=\"\" class=\"wp-image-908\" srcset=\"https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/74bd55e5-5cc6-49e8-be21-cf5c4a66042d.png 1028w, https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/74bd55e5-5cc6-49e8-be21-cf5c4a66042d-300x210.png 300w, https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/74bd55e5-5cc6-49e8-be21-cf5c4a66042d-1024x715.png 1024w, https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/74bd55e5-5cc6-49e8-be21-cf5c4a66042d-768x536.png 768w, https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/74bd55e5-5cc6-49e8-be21-cf5c4a66042d-18x12.png 18w\" sizes=\"auto, (max-width: 1028px) 100vw, 1028px\" \/><\/figure>\n\n\n\n<p><strong>ShareGPT-4o-Image forbedrer billedgenereringsydeevnen.<\/strong> Ved at finjustere Janus-Pro med ShareGPT-4o-Image genererede vi Janus-4o, som viser en betydeligt forbedret billedgenereringsydelse. Janus-4o underst\u00f8tter ogs\u00e5 tekst-til-billede og billede-til-billede-generering og overg\u00e5r andre benchmarks med kun 91.000 tr\u00e6ningspr\u00f8ver.<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"370\" src=\"https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/fc3b163f-d1d2-42f5-81bc-884eb677ea52.png\" alt=\"\" class=\"wp-image-910\" srcset=\"https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/fc3b163f-d1d2-42f5-81bc-884eb677ea52.png 1024w, https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/fc3b163f-d1d2-42f5-81bc-884eb677ea52-300x108.png 300w, https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/fc3b163f-d1d2-42f5-81bc-884eb677ea52-768x278.png 768w, https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/fc3b163f-d1d2-42f5-81bc-884eb677ea52-18x7.png 18w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n\n\n\n<p><strong>Oversigt over Janus-4o-modellen.<\/strong> Modellen er baseret p\u00e5 Janus-Pro og konstrueret ved at finjustere den p\u00e5 ShareGPT-4o-Image. Den inkorporerer forbedringer, der underst\u00f8tter tekst-til-billede og billede-til-billede-generering. B\u00e5de tekst-til-billede og tekst-til-billede-opgaver tr\u00e6nes i f\u00e6llesskab.<\/p>\n\n\n\n<figure class=\"wp-block-image\"><img loading=\"lazy\" decoding=\"async\" width=\"1058\" height=\"304\" src=\"https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/2b81408d-3c8b-45a8-ac73-ee0a48164c05.png\" alt=\"\" class=\"wp-image-909\" srcset=\"https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/2b81408d-3c8b-45a8-ac73-ee0a48164c05.png 1058w, https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/2b81408d-3c8b-45a8-ac73-ee0a48164c05-300x86.png 300w, https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/2b81408d-3c8b-45a8-ac73-ee0a48164c05-1024x294.png 1024w, https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/2b81408d-3c8b-45a8-ac73-ee0a48164c05-768x221.png 768w, https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/2b81408d-3c8b-45a8-ac73-ee0a48164c05-18x5.png 18w\" sizes=\"auto, (max-width: 1058px) 100vw, 1058px\" \/><\/figure>\n\n\n\n<h2 class=\"wp-block-heading\"><span class=\"ez-toc-section\" id=\"Experimental_Results\"><\/span>Eksperimentelle resultater<span class=\"ez-toc-section-end\"><\/span><\/h2>\n\n\n\n<figure class=\"wp-block-image\"><img loading=\"lazy\" decoding=\"async\" width=\"1072\" height=\"1140\" src=\"https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/72720ada-7418-4979-a8fd-4ce09050d696.png\" alt=\"\" class=\"wp-image-907\" srcset=\"https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/72720ada-7418-4979-a8fd-4ce09050d696.png 1072w, https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/72720ada-7418-4979-a8fd-4ce09050d696-282x300.png 282w, https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/72720ada-7418-4979-a8fd-4ce09050d696-963x1024.png 963w, https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/72720ada-7418-4979-a8fd-4ce09050d696-768x817.png 768w, https:\/\/janusai.pro\/wp-content\/uploads\/2025\/07\/72720ada-7418-4979-a8fd-4ce09050d696-11x12.png 11w\" sizes=\"auto, (max-width: 1072px) 100vw, 1072px\" \/><\/figure>\n\n\n\n<h2 class=\"wp-block-heading\"><span class=\"ez-toc-section\" id=\"Conclusions\"><\/span>Konklusioner<span class=\"ez-toc-section-end\"><\/span><\/h2>\n\n\n\n<p>ShareGPT-4o-Image er det f\u00f8rste datas\u00e6t i stor skala, der er i stand til at indfange GPT-4os avancerede billedgenereringsfunktioner inden for tekst-til-billede og tekst-til-billede-generering. Baseret p\u00e5 dette datas\u00e6t udviklede artiklen Janus-4o, en maskinl\u00e6ringsmodel (MLLM), der er i stand til at generere billeder i h\u00f8j kvalitet fra ren tekst eller billed-tekst-kombinationer.<\/p>\n\n\n\n<p>Janus-4o opn\u00e5r betydelige forbedringer i tekst-til-billede-generering og opn\u00e5r yderst konkurrencedygtige resultater i tekst-til-billede-opgaver, hvilket demonstrerer den h\u00f8je kvalitet og praktiske anvendelighed af ShareGPT-4o-Image.<\/p>\n\n\n\n<p>Takket v\u00e6re effektiviteten af selvregressiv billedgenerering baseret p\u00e5 MLLM kan Janus-4o tr\u00e6nes p\u00e5 bare 6 timer p\u00e5 en 8\u00d7A800 GPU-maskine og opn\u00e5r betydelige ydeevneforbedringer med ekstremt lave beregningskrav.<\/p>","protected":false},"excerpt":{"rendered":"<p>ShareGPT-4o-Image er et storstilet datas\u00e6t til billedgenerering af h\u00f8j kvalitet, hvor alle billeder genereres ved hj\u00e6lp af GPT-4os billedgenereringsfunktioner. Dette datas\u00e6t har til form\u00e5l at kombinere fordelene ved open source multimodale modeller med GPT-4os styrker inden for skabelse af visuelt indhold. Det inkluderer 45.000 tekst-til-billede og 46.000 billede-til-tekst-eksempler, hvilket g\u00f8r det til en praktisk ressource til forbedring af multimodale modeller i billed\u2026<\/p>","protected":false},"author":2,"featured_media":859,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"_kadence_starter_templates_imported_post":false,"_kad_post_transparent":"","_kad_post_title":"","_kad_post_layout":"","_kad_post_sidebar_id":"","_kad_post_content_style":"","_kad_post_vertical_padding":"","_kad_post_feature":"","_kad_post_feature_position":"","_kad_post_header":false,"_kad_post_footer":false,"footnotes":""},"categories":[1],"tags":[],"class_list":["post-906","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-uncategorized"],"_links":{"self":[{"href":"https:\/\/janusai.pro\/da\/wp-json\/wp\/v2\/posts\/906","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/janusai.pro\/da\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/janusai.pro\/da\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/janusai.pro\/da\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/janusai.pro\/da\/wp-json\/wp\/v2\/comments?post=906"}],"version-history":[{"count":1,"href":"https:\/\/janusai.pro\/da\/wp-json\/wp\/v2\/posts\/906\/revisions"}],"predecessor-version":[{"id":912,"href":"https:\/\/janusai.pro\/da\/wp-json\/wp\/v2\/posts\/906\/revisions\/912"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/janusai.pro\/da\/wp-json\/wp\/v2\/media\/859"}],"wp:attachment":[{"href":"https:\/\/janusai.pro\/da\/wp-json\/wp\/v2\/media?parent=906"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/janusai.pro\/da\/wp-json\/wp\/v2\/categories?post=906"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/janusai.pro\/da\/wp-json\/wp\/v2\/tags?post=906"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}