{"id":24509,"date":"2018-04-29T22:38:54","date_gmt":"2018-04-29T19:38:54","guid":{"rendered":"http:\/\/www.revolvia.com\/?post_type=mmg_articles&#038;p=24509"},"modified":"2018-04-29T22:38:54","modified_gmt":"2018-04-29T19:38:54","slug":"nvidianin-yeni-yapay-zekasi-kismen-silinmis-gorselleri-onariyor","status":"publish","type":"mmg_articles","link":"https:\/\/www.revolvia.com\/en\/mmg-article\/nvidianin-yeni-yapay-zekasi-kismen-silinmis-gorselleri-onariyor\/","title":{"rendered":"Nvidia\u2019n\u0131n yeni yapay zekas\u0131 k\u0131smen silinmi\u015f g\u00f6rselleri onar\u0131yor"},"content":{"rendered":"<p style=\"text-align: center;\">Yapay zeka nelere kadir de\u011fil mi? S\u00fcrekli olarak \u015fa\u015f\u0131rtacak, a\u011fz\u0131m\u0131z\u0131 a\u00e7\u0131k b\u0131rakacak \u015feyler ortaya \u00e7\u0131kmas\u0131na \u00f6n ayak oluyor. Tabii yapay zeka dedi\u011fimiz \u015fey asl\u0131nda insanl\u0131k olarak ilk etapta uyum sa\u011flayamad\u0131\u011f\u0131m\u0131z pek \u00e7ok teknolojiye kar\u015f\u0131l\u0131k geliyor. Eskiden insans\u0131 robotlar i\u00e7in yapay zeka tabirini kullan\u0131rken \u015fu anda makine \u00f6\u011frenme ve derin \u00f6\u011frenme s\u00fcre\u00e7lerine yapay zeka diyoruz. Anlayamad\u0131\u011f\u0131m\u0131z \u015feyi bu \u015fekilde tan\u0131mlamak \u00e7ok daha kolay elbette. Ancak anlayabilen ki\u015filer bunun \u00fczerinde \u00e7al\u0131\u015fmaya ve iyile\u015ftirmeye devam ediyor. Nvidia\u2019da bu markalardan birisi.\u00a0Guilin Liu liderli\u011findeki ara\u015ft\u0131rmac\u0131lar derin \u00f6\u011frenmeden faydalanarak g\u00f6rsellerdeki eksik par\u00e7alar\u0131 onaran bir ara\u00e7\u00a0<a href=\"https:\/\/arxiv.org\/abs\/1804.07723\" target=\"_blank\" rel=\"noopener\">geli\u015ftirdi<\/a>.<\/p>\n<p style=\"text-align: center;\"><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-102911\" src=\"https:\/\/bigumigu.com\/wp-content\/uploads\/2018\/04\/nvidia_ai_yapay_zeka_bigumigu_1.gif\" alt=\"\" width=\"720\" height=\"405\" \/><\/p>\n<h2 style=\"text-align: center;\">G\u00f6rsellerdeki bo\u015fluklar\u0131n\u0131z \u00f6zenle doldurulur hemen teslim edilir<\/h2>\n<p style=\"text-align: center;\">Nvidia\u2019n\u0131n yay\u0131nlad\u0131\u011f\u0131 tan\u0131t\u0131m videosunda, \u00f6nce sa\u011flam g\u00f6rsel al\u0131n\u0131yor ve rastgele noktalar tamamen siliniyor. Yapay zeka arac\u0131 ise bu bo\u015fluklar\u0131 g\u00f6rselin kalan\u0131yla tutarl\u0131 olacak \u015fekilde tamaml\u0131yor. Sonu\u00e7lar \u015fu anda m\u00fckemmel de\u011fil, ancak ziyadesiyle tutarl\u0131. Nvidia bu ara\u00e7 i\u00e7in ge\u00e7mi\u015fteki benzer teknolojilerden farkl\u0131 bir yol izleyerek geli\u015ftirme yapm\u0131\u015f. Bu sayede d\u00fczenlenen alan ile orijinal aras\u0131nda ton fark\u0131, beyaz \u015ferit veya renk kaymas\u0131 olmuyor. Bunun sebebi ise Nvidia\u2019n\u0131n algoritmas\u0131n\u0131n g\u00f6rselin merkezinden itibaren g\u00f6rseli alanlara ay\u0131rmak yerine bir b\u00fct\u00fcn olarak bakarak eksik b\u00f6lgeleri desen ve b\u00fct\u00fcnl\u00fck a\u00e7\u0131s\u0131ndan ele alarak doldurmas\u0131.<\/p>\n<p style=\"text-align: center;\">Adobe\u2019nin de benzer bir ara\u00e7 yay\u0131nlad\u0131\u011f\u0131n\u0131 ve hemen hemen ayn\u0131 \u015feyi yapt\u0131\u011f\u0131n\u0131 s\u00f6yleyebilirsiniz. Keza \u015firket Adobe Max 2017\u2019de SceneStitch ve\u00a0Cloak ara\u00e7lar\u0131n\u0131\u00a0<a href=\"https:\/\/bigumigu.com\/haber\/adobe-yeni-ozellikleriyle-gerceklige-meydan-okuyor\/\" target=\"_blank\" rel=\"noopener\">duyurmu\u015ftu<\/a>. Fakat bu ara\u00e7larla Nvidia\u2019n\u0131nki aras\u0131ndaki b\u00fcy\u00fck fark, ge\u00e7mi\u015f tekniklerde g\u00f6rsel yo\u011fun bir post prod\u00fcksiyon s\u00fcrecine tabii tutuluyordu. Bu i\u015flemle beraber ancak istenilen sonu\u00e7 al\u0131nabiliyordu. Post prod\u00fcksiyonunun gereklili\u011fi de g\u00f6rsellerin dikd\u00f6rtgenlere ayr\u0131larak b\u00f6lgesel incelenmesinden kaynaklan\u0131yor.<\/p>\n<p style=\"text-align: center;\">Arac\u0131n hen\u00fcz t\u00fcketiciye y\u00f6nelik sunulan bir ismi yok. \u00c7al\u0131\u015fma akademik seviyede. Fakat yapay zeka, derin \u00f6\u011frenme ve algoritmanlar\u0131n d\u00fcnyas\u0131nda g\u00f6rsellerin nas\u0131l manip\u00fcle edilebilece\u011fine tan\u0131kl\u0131k ediyoruz. Bu arac\u0131n da yeni ufuklar a\u00e7aca\u011f\u0131 kolayl\u0131kla \u00f6ng\u00f6r\u00fclebiliyor.<\/p>\n<p style=\"text-align: center;\"><img loading=\"lazy\" decoding=\"async\" class=\"alignnone size-full wp-image-102912\" src=\"https:\/\/bigumigu.com\/wp-content\/uploads\/2018\/04\/nvidia_ai_yapay_zeka_bigumigu_2.gif\" alt=\"\" width=\"720\" height=\"405\" \/><\/p>\n<div class=\"responsive-container\" style=\"text-align: center;\"><iframe loading=\"lazy\" src=\"https:\/\/www.youtube.com\/embed\/gg0F5JjKmhA?feature=oembed\" width=\"753\" height=\"424\" frameborder=\"0\" allowfullscreen=\"allowfullscreen\" data-mce-fragment=\"1\"><\/iframe><\/div>\n<p style=\"text-align: center;\"><em>G\u00f6rsel:\u00a0<a href=\"https:\/\/news.developer.nvidia.com\/new-ai-imaging-technique-reconstructs-photos-with-realistic-results\/?ncid=nv-twi-37107\" target=\"_blank\" rel=\"noopener\">Nvidia<\/a>, YouTube<\/em><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Nvidia&#8217;dan ara\u015ft\u0131rmac\u0131 Guilin Liu \u00f6ndeli\u011findeki ekip geli\u015ftirdi\u011fi ara\u00e7la derin \u00f6\u011frenme kullanarak g\u00f6rsellerde silinen alanlar\u0131n yeni bir y\u00f6ntemle onar\u0131lmas\u0131n\u0131 sa\u011fl\u0131yor.<\/p>\n","protected":false},"author":13,"featured_media":24510,"template":"","meta":{"_acf_changed":false,"footnotes":""},"tags":[439,358,222],"class_list":["post-24509","mmg_articles","type-mmg_articles","status-publish","has-post-thumbnail","hentry","tag-gorsel-onarma","tag-nvidia","tag-yapay-zeka"],"translation":{"provider":"WPGlobus","version":"3.0.0","language":"en","enabled_languages":["tr","en"],"languages":{"tr":{"title":true,"content":true,"excerpt":true},"en":{"title":false,"content":false,"excerpt":false}}},"acf":[],"_links":{"self":[{"href":"https:\/\/www.revolvia.com\/en\/wp-json\/wp\/v2\/mmg_articles\/24509","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.revolvia.com\/en\/wp-json\/wp\/v2\/mmg_articles"}],"about":[{"href":"https:\/\/www.revolvia.com\/en\/wp-json\/wp\/v2\/types\/mmg_articles"}],"author":[{"embeddable":true,"href":"https:\/\/www.revolvia.com\/en\/wp-json\/wp\/v2\/users\/13"}],"version-history":[{"count":0,"href":"https:\/\/www.revolvia.com\/en\/wp-json\/wp\/v2\/mmg_articles\/24509\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.revolvia.com\/en\/wp-json\/wp\/v2\/media\/24510"}],"wp:attachment":[{"href":"https:\/\/www.revolvia.com\/en\/wp-json\/wp\/v2\/media?parent=24509"}],"wp:term":[{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.revolvia.com\/en\/wp-json\/wp\/v2\/tags?post=24509"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}