[{"@context":"https:\/\/schema.org\/","@type":"Article","@id":"https:\/\/www.protonix.sk\/neuronova-siet-nvidia-dokaze-vratit-chybajuce-podrobnosti-fotografii\/#Article","mainEntityOfPage":"https:\/\/www.protonix.sk\/neuronova-siet-nvidia-dokaze-vratit-chybajuce-podrobnosti-fotografii\/","headline":"Neur\u00f3nov\u00e1 sie\u0165 NVIDIA dok\u00e1\u017ee vr\u00e1ti\u0165 ch\u00fdbaj\u00face podrobnosti fotografii","name":"Neur\u00f3nov\u00e1 sie\u0165 NVIDIA dok\u00e1\u017ee vr\u00e1ti\u0165 ch\u00fdbaj\u00face podrobnosti fotografii","description":"Technol\u00f3gia NVIDIA v s\u00fa\u010dasnosti nem\u00e1 vo svete \u017eiadne anal\u00f3gie. Algoritmus bol vyvinut\u00fd a predstaven\u00fd in\u017einierom Guilinom Liu. NVIDIA neur\u00f3nov\u00e9 siete teraz maj\u00fa konkurenciu v osobe spolo\u010dnosti Adobe, ktor\u00e1 pracuje na podobn\u00fdch technol\u00f3gi\u00e1ch. Technol\u00f3gia funguje v\u0161ak inak. Tvorcovia aplik\u00e1cie Photoshop pou\u017e\u00edvaj\u00fa AI, ktor\u00e9 po\u010das spracovania obrazu „po\u017ei\u010daj\u00fa“ kontextov\u00e9 mapovanie zo susedn\u00fdch pixelov a potom dokon\u010dia obraz","datePublished":"2025-02-26","dateModified":"2023-04-25","author":{"@type":"Person","@id":"https:\/\/www.protonix.sk\/author\/#Person","name":"protonix.sk\n","url":"https:\/\/www.protonix.sk\/author\/","identifier":1,"image":{"@type":"ImageObject","@id":"https:\/\/secure.gravatar.com\/avatar\/93869824c8aa580f0d16049ad31e69846ffc925e0cbb2826cc9554f53943da69?s=96&d=mm&r=g","url":"https:\/\/secure.gravatar.com\/avatar\/93869824c8aa580f0d16049ad31e69846ffc925e0cbb2826cc9554f53943da69?s=96&d=mm&r=g","height":96,"width":96}},"publisher":{"@type":"Organization","name":"protonix.sk","logo":{"@type":"ImageObject","@id":"\/logo.png","url":"\/logo.png","width":600,"height":60}},"image":{"@type":"ImageObject","@id":"https:\/\/www.protonix.sk\/wp-content\/uploads\/img_a287639_w2005_t1525600849.jpg","url":"https:\/\/www.protonix.sk\/wp-content\/uploads\/img_a287639_w2005_t1525600849.jpg","height":0,"width":0},"url":"https:\/\/www.protonix.sk\/neuronova-siet-nvidia-dokaze-vratit-chybajuce-podrobnosti-fotografii\/","about":["Technika"],"wordCount":395,"articleBody":"Technol\u00f3gia NVIDIA v s\u00fa\u010dasnosti nem\u00e1 vo svete \u017eiadne anal\u00f3gie. Algoritmus bol vyvinut\u00fd a predstaven\u00fd in\u017einierom Guilinom Liu. NVIDIA neur\u00f3nov\u00e9 siete teraz maj\u00fa konkurenciu v osobe spolo\u010dnosti Adobe, ktor\u00e1 pracuje na podobn\u00fdch technol\u00f3gi\u00e1ch. Technol\u00f3gia funguje v\u0161ak inak.Tvorcovia aplik\u00e1cie Photoshop pou\u017e\u00edvaj\u00fa AI, ktor\u00e9 po\u010das spracovania obrazu „po\u017ei\u010daj\u00fa“ kontextov\u00e9 mapovanie zo susedn\u00fdch pixelov a potom dokon\u010dia obraz v s\u00falade s t\u00fdmto. Tento pr\u00edstup nie je v\u017edy vhodn\u00fd, preto\u017ee ch\u00fdbaj\u00faci fragment m\u00f4\u017ee obsahova\u0165 nieko\u013eko farieb. Napr\u00edklad na okrajoch bude \u010derven\u00e1 a v strede – modr\u00e1. Syst\u00e9m vid\u00ed len okraj objektu a na z\u00e1klade toho „zaplav\u00ed“ v\u0161etko s \u010dervenou farbou.Algoritmus od spolo\u010dnosti NVIDIA funguje inak. Analyzuje kontext nie len \u010dasti oblasti, ale cel\u00e9ho obrazu ako celku, ktor\u00fd ur\u010duje vzory a vz\u00e1jomn\u00e9 vz\u0165ahy medzi objektmi. V\u0161etko je tak, aby v\u00fdsledok \u010do najviac pasoval nielen vo farebnosti, ale aj v s\u00favislosti s p\u00f4vodnou maa\u013ebou.Ke\u010f je technol\u00f3gia pripraven\u00e1, jej pou\u017eitie v\u00e1m umo\u017en\u00ed obnovi\u0165 fotografie za p\u00e1r min\u00fat.Pre prenos \u010dast\u00ed medzi obr\u00e1zkami (preklad obr\u00e1zkov do obr\u00e1zkov), ako aj v in\u00fdch aplik\u00e1ci\u00e1ch strojov\u00e9ho u\u010denia, sa pou\u017e\u00edvaj\u00fa r\u00f4zne pr\u00edstupy, ktor\u00e9 sa daj\u00fa rozdeli\u0165 do dvoch ve\u013ek\u00fdch typov:\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 vyu\u010dovanie s u\u010dite\u013eom\u00b7\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0\u00a0 bez u\u010dite\u013ea.V prvom pr\u00edpade s\u00fa \u00fadaje pre \u0161kolenie p\u00e1r po\u010diato\u010dn\u00fdch a po\u017eadovan\u00fdch \u00fadajov, tak\u017ee sa d\u00e1 algoritmus \u00a0s pr\u00edkladmi r\u00fdchlo nau\u010di\u0165. Probl\u00e9m s t\u00fdmto pr\u00edstupom je, \u017ee pri v\u00fdcviku je potrebn\u00e9 vytvori\u0165 ve\u013ek\u00e9 mno\u017einy d\u00e1tov\u00fdch p\u00e1rov, napr\u00edklad zdrojov\u00e9 a spracovan\u00e9 obrazy. V mnoh\u00fdch pr\u00edpadoch to vy\u017eaduje ve\u013ea \u010dasu a pe\u0148az\u00ed. \u0160pecialisti NVIDIA sa rozhodli pou\u017ei\u0165 druh\u00fd pr\u00edstup a pou\u017e\u00edvali \u0161kolenie bez u\u010dite\u013ea pre algoritmus prenosu obrazu. Je zalo\u017een\u00e1 na\u00a0 generat\u00edvnych a kontroverzn\u00fdch neur\u00f3nov\u00fdch sie\u0165ach . Tieto siete pozost\u00e1vaj\u00fa z dvoch konkuren\u010dn\u00fdch siet\u00ed. V tomto pr\u00edpade jeden z nich vytv\u00e1ra obraz a druh\u00fd ho odhaduje a ur\u010duje, \u010di je skuto\u010dn\u00fd alebo falo\u0161n\u00fd.\u00a0                                                                                                                                                                                                                                                                                                                                                                                        4.4\/5 - (8 votes)        "},{"@context":"https:\/\/schema.org\/","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Neur\u00f3nov\u00e1 sie\u0165 NVIDIA dok\u00e1\u017ee vr\u00e1ti\u0165 ch\u00fdbaj\u00face podrobnosti fotografii","item":"https:\/\/www.protonix.sk\/neuronova-siet-nvidia-dokaze-vratit-chybajuce-podrobnosti-fotografii\/#breadcrumbitem"}]}]