{"id":16477,"date":"2024-08-09T09:02:07","date_gmt":"2024-08-09T07:02:07","guid":{"rendered":"https:\/\/www.sling.si\/?post_type=wpdmpro&#038;p=16477"},"modified":"2025-07-16T14:49:00","modified_gmt":"2025-07-16T12:49:00","slug":"super-resolucija","status":"publish","type":"wpdmpro","link":"https:\/\/www.sling.si\/en\/download\/super-resolucija\/","title":{"rendered":"Super-resolucija"},"content":{"rendered":"<p>\u0160tevilni preboji na podro\u010dju hitrosti in natan\u010dnosti super-resolucije posameznih slik (ang. single image super-resolution, SISR) so bili \u017ee dose\u017eeni. Eden od najve\u010djih \u0161e nere\u0161enih izzivov je, kako obnoviti fine teksturne podrobnosti pri uporabi super-resolucije z velikimi faktorji pove\u010danja. Tipi\u010dna re\u0161itev za SISR vklju\u010duje uporabo konvolucijske nevronske mre\u017ee (ang. convolutional neural network, CNN), vendar pa so na voljo tudi novi pristopi, ki uporabljajo generativne nasprotni\u0161ke mre\u017ee (ang. generative adversarial network, GAN). Obna\u0161anje metod super-resolucije, ki temeljijo na optimizacij, je v principu vodeno in pogojeno z izbiro funkcije izgube (angl. loss function). V tem delu predstavljamo opis in evalvacijo mre\u017e SRResNet in SRGAN. SRResNet predstavlja rezidualno globoko mre\u017eo, SRGAN pa je generativna nasprotni\u0161ka mre\u017ea. Obe mre\u017ei uporabimo za re\u0161evanje problema super-resolucije slik (angl. super-resolution, SR). SRResNet je zmo\u017ena iz mo\u010dno pomanj\u0161anih slik obnoviti foto-realisti\u010dne slike sprejemljive kakovosti. SRGAN je zmo\u017ena ekstrapolirati foto-realisti\u010dne naravne slike z visokimi faktorji pove\u010danja. To dose\u017eemo z uporabo funkcije zaznavne izgube (angl. perceptual loss function), ki je sestavljena iz nasprotni\u0161ke izgube (angl. adversarial loss) in vsebinske izgube (angl. content loss). Nasprotni\u0161ka izguba potisne re\u0161itev na razse\u017enost naravnih slik z uporabo diskriminatorne mre\u017ee. Ta je nau\u010dena razlikovati med originalnimi foto-realisti\u010dnimi slikami in pove\u010danimi slikami. Vsebinska izguba pa je motivirana z zaznavno podobnostjo namesto s podobnostjo v prostoru pikslov.<\/p>\n<p>Prenesi celotno besedilo:<\/p>\n<p><a href=\"https:\/\/www.sling.si?wpdmdl=16477&amp;ind=1726233067268\"><img loading=\"lazy\" decoding=\"async\" class=\"fy-content-image fy-lazy js-lazy alignnone\" src=\"data:image\/svg+xml,%3Csvg%20width%3D%2280%22%20height%3D%2280%22%20xmlns%3D%22http:\/\/www.w3.org\/2000\/svg%22%20viewBox%3D%220%200%2080%2080%22%3E%3C\/svg%3E\" alt=\"Prenesi prispevek\" width=\"80\" height=\"80\" data-src=\"https:\/\/www.sling.si\/wp-content\/plugins\/download-manager\/assets\/file-type-icons\/resume-download.png\"><div class=\"fy-image-loading fy-image-loading--spinner\" aria-hidden=\"true\"><\/div><\/a><\/p>\n<p>\u010clanek je v celoti na voljo le v angle\u0161kem jeziku.<\/p>\n<p>Avtor: Sebastien Strban; Univerza v Ljubljani<\/p>\n","protected":false},"excerpt":{"rendered":"<p>\u0160tevilni preboji na podro\u010dju hitrosti in natan\u010dnosti super-resolucije posameznih slik (ang. single image super-resolution, SISR) so bili \u017ee dose\u017eeni. Eden od najve\u010djih \u0161e nere\u0161enih izzivov je, kako obnoviti fine teksturne podrobnosti pri uporabi super-resolucije z velikimi faktorji pove\u010danja.<\/p>\n","protected":false},"author":13,"featured_media":15540,"comment_status":"closed","ping_status":"closed","template":"","meta":{"_acf_changed":false,"__wpdm_changelog":[]},"wpdmcategory":[630],"wpdmtag":[485,633,632,634,631,483,657],"class_list":["post-16477","wpdmpro","type-wpdmpro","status-publish","has-post-thumbnail","hentry","wpdmcategory-strokovni-clanek","wpdmtag-fri-ul","wpdmtag-generative-adversarial-network","wpdmtag-sisr","wpdmtag-srgan","wpdmtag-super-resolucija","wpdmtag-superracunalnistvo","wpdmtag-ul"],"acf":[],"aioseo_notices":[],"_links":{"self":[{"href":"https:\/\/www.sling.si\/en\/wp-json\/wp\/v2\/wpdmpro\/16477","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.sling.si\/en\/wp-json\/wp\/v2\/wpdmpro"}],"about":[{"href":"https:\/\/www.sling.si\/en\/wp-json\/wp\/v2\/types\/wpdmpro"}],"author":[{"embeddable":true,"href":"https:\/\/www.sling.si\/en\/wp-json\/wp\/v2\/users\/13"}],"replies":[{"embeddable":true,"href":"https:\/\/www.sling.si\/en\/wp-json\/wp\/v2\/comments?post=16477"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.sling.si\/en\/wp-json\/wp\/v2\/media\/15540"}],"wp:attachment":[{"href":"https:\/\/www.sling.si\/en\/wp-json\/wp\/v2\/media?parent=16477"}],"wp:term":[{"taxonomy":"wpdmcategory","embeddable":true,"href":"https:\/\/www.sling.si\/en\/wp-json\/wp\/v2\/wpdmcategory?post=16477"},{"taxonomy":"wpdmtag","embeddable":true,"href":"https:\/\/www.sling.si\/en\/wp-json\/wp\/v2\/wpdmtag?post=16477"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}