{"id":54780,"date":"2020-01-04T00:00:00","date_gmt":"2020-01-03T21:00:00","guid":{"rendered":"https:\/\/prohoster.info\/blog\/blog_prohoster\/nejroseti-kuda-eto-vse-dvizhetsya"},"modified":"2020-02-18T14:02:50","modified_gmt":"2020-02-18T11:02:50","slug":"nejroseti-kuda-eto-vse-dvizhetsya","status":"publish","type":"post","link":"https:\/\/prohoster.info\/hu\/blog\/news\/nejroseti-kuda-eto-vse-dvizhetsya","title":{"rendered":"Neur\u00e1lis h\u00e1l\u00f3zatok. Merre halad ez az eg\u00e9sz?","gt_translate_keys":[{"key":"rendered","format":"text"}]},"content":{"rendered":"<p>A cikk k\u00e9t r\u00e9szb\u0151l \u00e1ll:<\/p>\n<p><\/p>\n<ol>\n<li>R\u00f6vid le\u00edr\u00e1s n\u00e9h\u00e1ny architekt\u00far\u00e1r\u00f3l, amely a k\u00e9pek objektumfelismer\u00e9s\u00e9re \u00e9s a k\u00e9pek szegment\u00e1l\u00e1s\u00e1ra vonatkozik, a sz\u00e1momra leg\u00e9rthet\u0151bb hivatkoz\u00e1sokkal. Megpr\u00f3b\u00e1ltam vide\u00f3magyar\u00e1zatokat v\u00e1lasztani, lehet\u0151leg orosz nyelven.<\/li>\n<li>A m\u00e1sodik r\u00e9sz azt pr\u00f3b\u00e1lja meg\u00e9rteni, hogy merre fejl\u0151dnek a neur\u00e1lis h\u00e1l\u00f3zatok architekt\u00far\u00e1i \u00e9s a r\u00e1juk \u00e9p\u00fcl\u0151 technol\u00f3gi\u00e1k.<\/li>\n<\/ol>\n<p><\/p>\n<p><img decoding=\"async\" alt=\"Neur\u00e1lis h\u00e1l\u00f3zatok. Merre halad ez az eg\u00e9sz?\" src=\"\/wp-content\/uploads\/2020\/01\/3e0238547dc956dbb069c11241e4534f.png\" style=\"display:block;margin: 0 auto;\" \/><\/p>\n<p><\/p>\n<p>1. \u00e1bra \u2013 A neur\u00e1lis h\u00e1l\u00f3zatok architekt\u00far\u00e1inak meg\u00e9rt\u00e9se nem egyszer\u0171<\/p>\n<p><\/p>\n<p>Minden azzal kezd\u0151d\u00f6tt, hogy k\u00e9t bemutat\u00f3alkalmaz\u00e1st k\u00e9sz\u00edtettem k\u00e9poszt\u00e1lyoz\u00e1sra \u00e9s objektumfelismer\u00e9sre Android telefonon:<\/p>\n<p><\/p>\n<ul>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/foobar167\/junkyard\/tree\/master\/object_classifier\">Back-end demo<\/a><\/noindex>, amikor az adatok a szerveren ker\u00fclnek feldolgoz\u00e1sra \u00e9s \u00e1tker\u00fclnek a telefonra. H\u00e1romf\u00e9le medve k\u00e9poszt\u00e1lyoz\u00e1sa: barna, fekete \u00e9s pl\u00fcss.<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/foobar167\/android\/tree\/master\/object_detection_demo\">Front-end demo<\/a><\/noindex>, amikor az adatok mag\u00e1n a telefonon ker\u00fclnek feldolgoz\u00e1sra. H\u00e1romf\u00e9le objektumfelismer\u00e9s: mogyor\u00f3, f\u00fcge \u00e9s datolya.<\/li>\n<\/ul>\n<p><noindex><a rel=\"nofollow\" name=\"habracut\"><\/a><\/noindex><\/p>\n<p>K\u00fcl\u00f6nbs\u00e9g van a k\u00e9poszt\u00e1lyoz\u00e1si feladatok, a k\u00e9peken t\u00f6rt\u00e9n\u0151 objektumfelismer\u00e9s \u00e9s <noindex><a rel=\"nofollow\" href=\"https:\/\/medium.com\/analytics-vidhya\/image-classification-vs-object-detection-vs-image-segmentation-f36db85fe81\">a k\u00e9pek szegment\u00e1l\u00e1sa<\/a><\/noindex>. Ez\u00e9rt felmer\u00fclt az ig\u00e9ny, hogy megtudjam, mely neur\u00e1lis h\u00e1l\u00f3zati architekt\u00far\u00e1k ismerik fel az objektumokat a k\u00e9peken \u00e9s melyek tudj\u00e1k szegment\u00e1lni. A k\u00f6vetkez\u0151 architekt\u00far\u00e1k p\u00e9ld\u00e1it tal\u00e1ltam, a sz\u00e1momra leg\u00e9rthet\u0151bb hivatkoz\u00e1sokkal:<\/p>\n<p><\/p>\n<ul>\n<li>R-CNN alap\u00fa architekt\u00far\u00e1k sorozata (<strong>R<\/strong>egions with <strong>C<\/strong>onvolution <strong>N<\/strong>eural <strong>N<\/strong>etworks features): R-CNN, Fast R-CNN, <noindex><a rel=\"nofollow\" href=\"https:\/\/medium.com\/@smallfishbigsea\/faster-r-cnn-explained-864d4fb7e3f8\">Faster R-CNN<\/a><\/noindex>, <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/0vt05rQqk_I\">Mask R-CNN<\/a><\/noindex>. A k\u00e9peken t\u00f6rt\u00e9n\u0151 objektumfelismer\u00e9shez a Region Proposal Network (RPN) mechanizmus seg\u00edts\u00e9g\u00e9vel korl\u00e1tozott r\u00e9gi\u00f3kat (bounding boxes) hat\u00e1roznak meg. Kezdetben lassabb Selective Search mechanizmust alkalmaztak az RPN helyett. Azut\u00e1n a kiemelt korl\u00e1tozott r\u00e9gi\u00f3kat egy hagyom\u00e1nyos neur\u00e1lis h\u00e1l\u00f3zat bemenet\u00e9re k\u00fcldik oszt\u00e1lyoz\u00e1sra. Az R-CNN architekt\u00far\u00e1ban egy\u00e9rtelm\u0171 \"for\" ciklusok vannak az egyes korl\u00e1tozott r\u00e9gi\u00f3k \u00e1tvizsg\u00e1l\u00e1s\u00e1ra, \u00f6sszesen ak\u00e1r 2000 iter\u00e1ci\u00f3 az AlexNet bels\u0151 h\u00e1l\u00f3zaton. Az egy\u00e9rtelm\u0171 \"for\" ciklusok miatt cs\u00f6kken a k\u00e9pek feldolgoz\u00e1si sebess\u00e9ge. Minden \u00faj architekt\u00fara verzi\u00f3val cs\u00f6kken az egy\u00e9rtelm\u0171 ciklusok, a bels\u0151 neur\u00e1lis h\u00e1l\u00f3zaton val\u00f3 \u00e1tmenetek sz\u00e1ma, \u00e9s sok egy\u00e9b v\u00e1ltoz\u00e1st v\u00e9geznek a sebess\u00e9g n\u00f6vel\u00e9s\u00e9re, valamint a t\u00e1rgyak \u00e9szlel\u00e9s\u00e9nek feladat\u00e1t a Mask R-CNN-ben t\u00f6rt\u00e9n\u0151 objektum-szegment\u00e1ci\u00f3ra t\u00f6rt\u00e9n\u0151 \u00e1tkonvert\u00e1l\u00e1sra.<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/L0tzmv--CGY\">YOLO<\/a><\/noindex> (<strong>Y<\/strong>ou <strong>O<\/strong>nly <strong>L<\/strong>ook <strong>O<\/strong>A YOLO (You Only Look Once) \u2013 az els\u0151 neur\u00e1lis h\u00e1l\u00f3zat, amely val\u00f3s id\u0151ben k\u00e9pes azonos\u00edtani a t\u00e1rgyakat mobil eszk\u00f6z\u00f6k\u00f6n. K\u00fcl\u00f6nleges jellemz\u0151je: a t\u00e1rgyak egyetlen \u00e1tfut\u00e1ssal t\u00f6rt\u00e9n\u0151 megk\u00fcl\u00f6nb\u00f6ztet\u00e9se (el\u00e9g egyszer r\u00e1n\u00e9zni). Ez\u00e9rt a YOLO architekt\u00far\u00e1ban nincsenek kifejezett \u201efor\u201d ciklusok, ami gyors m\u0171k\u00f6d\u00e9st eredm\u00e9nyez. P\u00e9ld\u00e1ul, hasonl\u00f3 a NumPy-hoz, ahol a m\u00e1trix m\u0171veletek sor\u00e1n sincsenek kifejezett \u201efor\u201d ciklusok, ezeket a NumPy alacsonyabb szint\u0171 architekt\u00far\u00e1j\u00e1ban a C programoz\u00e1si nyelv val\u00f3s\u00edtja meg. A YOLO el\u0151re meghat\u00e1rozott ablakokb\u00f3l \u00e1ll\u00f3 r\u00e1csot haszn\u00e1l. Annak \u00e9rdek\u00e9ben, hogy ugyanazt a t\u00e1rgyat ne azonos\u00edts\u00e1k t\u00f6bbsz\u00f6r, \u00e1tfed\u00e9si egy\u00fctthat\u00f3t (IoU) alkalmaznak. <strong>I<\/strong>metsz\u0151d\u00e9s <strong>) k\u00f6z\u00f6tt. Ezut\u00e1n haszn\u00e1l egy oper\u00e1tort a jogosults\u00e1gok hozz\u00e1ad\u00e1s\u00e1ra vagy elt\u00e1vol\u00edt\u00e1s\u00e1ra az aktu\u00e1lis m\u00f3db\u00f3l, vagy abszol\u00fat jogokat \u00e1ll\u00edt be.<\/strong>uni\u00f3. <strong>U<\/strong>Ez az architekt\u00fara sz\u00e9les sk\u00e1l\u00e1n m\u0171k\u00f6dik \u00e9s magas <noindex><a rel=\"nofollow\" href=\"https:\/\/ru.wikipedia.org\/wiki\/%D0%A0%D0%BE%D0%B1%D0%B0%D1%81%D1%82%D0%BD%D0%BE%D1%81%D1%82%D1%8C\">robusztuss\u00e1ggal<\/a><\/noindex>: a modell f\u00e9nyk\u00e9peken tanulhat\u00f3, de j\u00f3l m\u0171k\u00f6dik rajzolt k\u00e9peken is.<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/P8e-G-Mhx4k\">SSD<\/a><\/noindex> (<strong>S<\/strong>Egys\u00e9ges <strong>S<\/strong>MultiBox Megel\u0151z\u0151 <strong>D<\/strong>detektor) \u2013 a YOLO architekt\u00far\u00e1j\u00e1nak legjobban siker\u00fclt \u201ehackjeit\u201d alkalmazz\u00e1k, p\u00e9ld\u00e1ul a nem maxim\u00e1lis visszaszor\u00edt\u00e1st, \u00e9s \u00fajakat adnak hozz\u00e1, hogy a neur\u00e1lis h\u00e1l\u00f3zat gyorsabban \u00e9s pontosabban m\u0171k\u00f6dj\u00f6n. K\u00fcl\u00f6nleges jellemz\u0151je: a t\u00e1rgyak megk\u00fcl\u00f6nb\u00f6ztet\u00e9se egy \u00e1tfut\u00e1ssal az el\u0151re meghat\u00e1rozott ablakok (default box) seg\u00edts\u00e9g\u00e9vel k\u00e9pkocka piramisokon. A k\u00e9pet piramisokban k\u00f3dolj\u00e1k konvol\u00faci\u00f3s tenzorokkal a sorozatos konvol\u00faci\u00f3s \u00e9s max-pooling m\u0171veletek sor\u00e1n (max-pooling m\u0171veletn\u00e9l a t\u00e9rbeli dimenzi\u00f3 cs\u00f6kken). \u00cdgy egyszerre hat\u00e1rozhat\u00f3k meg a nagy \u00e9s a kis t\u00e1rgyak.<\/li>\n<li>MobileSSD (<strong>Mobile<\/strong>NetV2 + <strong>SSD<\/strong>) \u2013 k\u00e9t neur\u00e1lis h\u00e1l\u00f3zat architekt\u00far\u00e1j\u00e1nak kombin\u00e1ci\u00f3ja. Az els\u0151 h\u00e1l\u00f3zat <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/post\/352804\/\">MobileNetV2<\/a><\/noindex> gyorsan m\u0171k\u00f6dik \u00e9s n\u00f6veli az azonos\u00edt\u00e1s pontoss\u00e1g\u00e1t. A MobileNetV2 a VGG-16-ot helyettes\u00edti, amelyet eredetileg haszn\u00e1ltak a <noindex><a rel=\"nofollow\" href=\"https:\/\/arxiv.org\/abs\/1512.02325\">eredeti cikkben<\/a><\/noindex>. A m\u00e1sodik h\u00e1l\u00f3zat az SSD hat\u00e1rozza meg a t\u00e1rgyak hely\u00e9t a k\u00e9pen.<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/ge_RT5wvHvY\">SqueezeNet<\/a><\/noindex> \u2013 nagyon kicsi, de pontos neur\u00e1lis h\u00e1l\u00f3zat. \u00d6nmag\u00e1ban nem oldja meg a t\u00e1rgyak azonos\u00edt\u00e1s\u00e1t. De k\u00fcl\u00f6nb\u00f6z\u0151 architekt\u00far\u00e1k kombin\u00e1l\u00e1s\u00e1val alkalmazhat\u00f3. Haszn\u00e1lhat\u00f3 mobil eszk\u00f6z\u00f6kben is. K\u00fcl\u00f6nleges jellemz\u0151je, hogy el\u0151sz\u00f6r az adatokat n\u00e9gy 1\u00d71-es konvol\u00faci\u00f3s sz\u0171r\u0151re t\u00f6m\u00f6r\u00edtik, majd n\u00e9gy 1\u00d71-es \u00e9s n\u00e9gy 3\u00d73-as konvol\u00faci\u00f3s sz\u0171r\u0151re b\u0151v\u00edtik. Ezt a t\u00f6m\u00f6r\u00edt\u00e9si-b\u0151v\u00edt\u00e9si iter\u00e1ci\u00f3t \u201eFire Module\u201d-nak nevezik.<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/b6jhopSMit8\">DeepLab<\/a><\/noindex> (A m\u00e9ly konvol\u00faci\u00f3s h\u00e1l\u00f3zatokkal v\u00e9gzett szemantikus k\u00e9pszegment\u00e1l\u00e1s) \u2013 objektumok szegment\u00e1l\u00e1sa a k\u00e9pen. Az architekt\u00fara jellegzetess\u00e9ge a ritk\u00edtott (dilated convolution) convol\u00faci\u00f3, amely meg\u0151rzi a t\u00e9rbeli felbont\u00e1st. Ezt k\u00f6veti a szegment\u00e1l\u00e1si eredm\u00e9nyek ut\u00f3munk\u00e1lata egy grafikus val\u00f3sz\u00edn\u0171s\u00e9gi modell (conditional random field) alkalmaz\u00e1s\u00e1val, amely elt\u00e1vol\u00edtja a kis zajokat a szegment\u00e1ci\u00f3b\u00f3l, \u00e9s jav\u00edtja a szegment\u00e1lt k\u00e9p min\u0151s\u00e9g\u00e9t. A \u201egrafikus val\u00f3sz\u00edn\u0171s\u00e9gi modell\u201d m\u00f6g\u00f6tt egy szok\u00e1sos Gauss-sz\u0171r\u0151 rejt\u0151zik, amelyet \u00f6t pont k\u00f6r\u00fcli k\u00f6zel\u00edt\u00e9ssel alak\u00edtottak ki.<\/li>\n<li>Pr\u00f3b\u00e1ltam meg\u00e9rteni a fel\u00e9p\u00edt\u00e9s\u00e9t <noindex><a rel=\"nofollow\" href=\"https:\/\/arxiv.org\/abs\/1711.06897\">RefineDet<\/a><\/noindex> (Single-Shot <strong>Refine<\/strong>ment Neur\u00e1lis H\u00e1l\u00f3zat az Objektum <strong>Det<\/strong>ekci\u00f3hoz), de alig \u00e9rtettem meg valamit.<\/li>\n<li>Azt is megn\u00e9ztem, hogyan m\u0171k\u00f6dik a figyelmi technol\u00f3gia (attention): <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/W2rWgXJBZhU\">video1<\/a><\/noindex>, <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/iDulhoQ2pro\">video2<\/a><\/noindex>, <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/H6Qiegq_36c\">video3<\/a><\/noindex>. Az architekt\u00fara \u201efigyelem\u201d saj\u00e1toss\u00e1ga a figyelem kiemelt r\u00e9gi\u00f3inak automatikus meghat\u00e1roz\u00e1sa a k\u00e9pen (RoI, <strong>R<\/strong>\u00e9rdekl\u0151d\u00e9si <strong>) k\u00f6z\u00f6tt. Ezut\u00e1n haszn\u00e1l egy oper\u00e1tort a jogosults\u00e1gok hozz\u00e1ad\u00e1s\u00e1ra vagy elt\u00e1vol\u00edt\u00e1s\u00e1ra az aktu\u00e1lis m\u00f3db\u00f3l, vagy abszol\u00fat jogokat \u00e1ll\u00edt be.<\/strong>f <strong>I<\/strong>egi\u00f3k) egy Attention Unit nev\u0171 neur\u00e1lis h\u00e1l\u00f3 seg\u00edts\u00e9g\u00e9vel. A kiemelt figyelmi r\u00e9gi\u00f3k hasonl\u00edtanak a korl\u00e1tozott r\u00e9gi\u00f3kra (bounding boxes), de ellent\u00e9tben azokkal nem r\u00f6gz\u00edtettek a k\u00e9pen, \u00e9s elmos\u00f3dott hat\u00e1rvonalakkal rendelkezhetnek. Ezut\u00e1n a kiemelt figyelmi r\u00e9gi\u00f3kb\u00f3l jellemz\u0151k (fich\u00e1k) ker\u00fclnek kiemel\u00e9sre, amelyeket visszaes\u0151 neur\u00e1lis h\u00e1l\u00f3zatokhoz \u201et\u00e1pl\u00e1lnak\u201d, olyan architekt\u00far\u00e1kkal, mint <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/5lUUrREboSk\">LSDM, GRU vagy Vanilla RNN<\/a><\/noindex>. A visszaes\u0151 neur\u00e1lis h\u00e1l\u00f3zatok k\u00e9pesek elemezni a jellemz\u0151k k\u00f6z\u00f6tti kapcsolatokat a sorozatban. A visszaes\u0151 neur\u00e1lis h\u00e1l\u00f3zatokat kezdetben m\u00e1s nyelvekre t\u00f6rt\u00e9n\u0151 sz\u00f6vegek ford\u00edt\u00e1s\u00e1ra alkalmazt\u00e1k, most pedig a <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/e-WB4lfg30M\">k\u00e9pek sz\u00f6vegg\u00e9 t\u00f6rt\u00e9n\u0151 ford\u00edt\u00e1s\u00e1ra<\/a><\/noindex> \u00e9s <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/rAbhypxs1qQ\">sz\u00f6vegek k\u00e9pp\u00e9 alak\u00edt\u00e1s\u00e1ra<\/a><\/noindex>.<\/li>\n<\/ul>\n<p><\/p>\n<p>A fenti architekt\u00far\u00e1k tanulm\u00e1nyoz\u00e1sa sor\u00e1n <strong>r\u00e1j\u00f6ttem, hogy semmit nem \u00e9rtek.<\/strong>. Nem arr\u00f3l van sz\u00f3, hogy a neur\u00e1lis h\u00e1l\u00f3mnak probl\u00e9m\u00e1i vannak a figyelem mechanizmus\u00e1val. Az \u00f6sszes architekt\u00fara l\u00e9trehoz\u00e1sa egy \u00f3ri\u00e1si hackathonra hasonl\u00edt, ahol a szerz\u0151k a hackekben versenyeznek. A hack (hack) egy neh\u00e9z programoz\u00e1si probl\u00e9ma gyors megold\u00e1sa. Teh\u00e1t ezek k\u00f6z\u00f6tt az architekt\u00far\u00e1k k\u00f6z\u00f6tt nincs l\u00e1that\u00f3 \u00e9s \u00e9rthet\u0151 logikai kapcsolat. Ami \u00f6sszek\u00f6ti \u0151ket, az a legjobb hackek gy\u0171jtem\u00e9nye, amelyeket egym\u00e1st\u00f3l k\u00f6lcs\u00f6n\u00f6znek, plusz az \u00f6sszes felhaszn\u00e1l\u00f3 sz\u00e1m\u00e1ra k\u00f6z\u00f6s <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/Ilg3gGewQ5U\">ford\u00edt\u00e1si m\u0171velet visszacsatol\u00e1ssal<\/a><\/noindex> (visszaterjed\u00e9s, backpropagation). Nincs <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/post\/272473\/\">rendszerszeml\u00e9let<\/a><\/noindex>! Nem vil\u00e1gos, hogy mit kell v\u00e1ltoztatni \u00e9s hogyan optimaliz\u00e1lni a megl\u00e9v\u0151 eredm\u00e9nyeket.<\/p>\n<p><\/p>\n<p>Azoknak a hackeknek a logikai \u00f6sszef\u00fcgg\u00e9s\u00e9nek hi\u00e1nya miatt rendk\u00edv\u00fcl neh\u00e9z \u0151ket megjegyezni \u00e9s a gyakorlatban alkalmazni. Ezek fragment\u00e1lt ismeretek. Legjobb esetben n\u00e9h\u00e1ny \u00e9rdekes \u00e9s v\u00e1ratlan pillanat megmarad, de a meg\u00e9rtett \u00e9s meg nem \u00e9rtett t\u00f6bbs\u00e9ge m\u00e1r n\u00e9h\u00e1ny nap ut\u00e1n elt\u0171nik az eml\u00e9kezetb\u0151l. J\u00f3 lenne, ha legal\u00e1bb az architekt\u00fara neve esz\u00e9be jutna egy h\u00e9t eltelt\u00e9vel. Pedig a cikkek olvas\u00e1s\u00e1ra \u00e9s a bemutat\u00f3 vide\u00f3k megtekint\u00e9s\u00e9re \u00f3r\u00e1kat, s\u0151t napokat k\u00f6lt\u00f6ttek!<\/p>\n<p><\/p>\n<p><img decoding=\"async\" alt=\"Neur\u00e1lis h\u00e1l\u00f3zatok. Merre halad ez az eg\u00e9sz?\" src=\"\/wp-content\/uploads\/2020\/01\/86f0f24e3be1d0f1a210f3132897d981.png\" style=\"display:block;margin: 0 auto;\" \/><\/p>\n<p><\/p>\n<p>2. \u00e1bra \u2013 <noindex><a rel=\"nofollow\" href=\"https:\/\/www.asimovinstitute.org\/neural-network-zoo\/\">A neur\u00e1lis h\u00e1l\u00f3zatok \u00e1llatkertje<\/a><\/noindex><\/p>\n<p><\/p>\n<p>Szem\u00e9lyes v\u00e9lem\u00e9nyem szerint a tudom\u00e1nyos cikkek szerz\u0151i mindent megtesznek annak \u00e9rdek\u00e9ben, hogy m\u00e9g ezek a fragment\u00e1lt ismeretek is \u00e9rthetetlenek maradjanak az olvas\u00f3 sz\u00e1m\u00e1ra. De a gerund form\u00e1k t\u00edz soros mondatokban, amelyek k\u00e9pletei a \u201ep\u00f3kh\u00e1l\u00f3b\u00f3l\u201d sz\u00e1rmaznak \u2013 ez egy m\u00e1sik cikk t\u00e9m\u00e1ja (probl\u00e9ma <noindex><a rel=\"nofollow\" href=\"https:\/\/en.wikipedia.org\/wiki\/Publish_or_perish\">publish or perish<\/a><\/noindex>).<\/p>\n<p><\/p>\n<p>Ez\u00e9rt v\u00e1lt sz\u00fcks\u00e9gess\u00e9 a neur\u00e1lis h\u00e1l\u00f3zatokkal kapcsolatos inform\u00e1ci\u00f3k rendszerez\u00e9se \u00e9s ez\u00e1ltal a meg\u00e9rt\u00e9s \u00e9s eml\u00e9kez\u00e9s min\u0151s\u00e9g\u00e9nek n\u00f6vel\u00e9se. Ez\u00e9rt az egyes mesters\u00e9ges neur\u00e1lis h\u00e1l\u00f3zatok technol\u00f3gi\u00e1inak \u00e9s architekt\u00far\u00e1inak elemz\u00e9s\u00e9nek f\u0151 t\u00e9m\u00e1j\u00e1v\u00e1 v\u00e1lt a k\u00f6vetkez\u0151 k\u00e9rd\u00e9s: <strong>tudni, hov\u00e1 halad ez az eg\u00e9sz<\/strong>, nem pedig egy konkr\u00e9t neur\u00e1lis h\u00e1l\u00f3zat fel\u00e9p\u00edt\u00e9se k\u00fcl\u00f6n-k\u00fcl\u00f6n.<\/p>\n<p><\/p>\n<p>Hov\u00e1 halad ez az eg\u00e9sz. F\u0151bb eredm\u00e9nyek:<\/p>\n<p><\/p>\n<ul>\n<li>Az elm\u00falt k\u00e9t \u00e9vben a g\u00e9pi tanul\u00e1s ter\u00fclet\u00e9n a startupok sz\u00e1ma <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/company\/recognitor\/blog\/455676\/\">dr\u00e1maian cs\u00f6kkent<\/a><\/noindex>. Lehets\u00e9ges oka: \u201ea neur\u00e1lis h\u00e1l\u00f3zatok m\u00e1r nem \u00fajdons\u00e1g.\u201d<\/li>\n<li>Mindenki l\u00e9trehozhat egy m\u0171k\u00f6d\u0151 neur\u00e1lis h\u00e1l\u00f3zatot egy egyszer\u0171 feladat megold\u00e1s\u00e1ra. Ehhez egy k\u00e9szen \u00e1ll\u00f3 modellt vesz a \u201emodellek \u00e1llatkertj\u00e9b\u0151l\u201d (model zoo) \u00e9s megtr\u00e9neli a neur\u00e1lis h\u00e1l\u00f3zat utols\u00f3 r\u00e9teg\u00e9t (<noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/yofjFQddwHE\">transfer learning<\/a><\/noindex>) a k\u00e9szen \u00e1ll\u00f3 adatokon kereszt\u00fcl <noindex><a rel=\"nofollow\" href=\"https:\/\/toolbox.google.com\/datasetsearch\">Google Dataset Search<\/a><\/noindex> vagy <noindex><a rel=\"nofollow\" href=\"https:\/\/www.kaggle.com\/datasets\">25 ezer Kaggle adatb\u00e1zisban<\/a><\/noindex> ingyenes <noindex><a rel=\"nofollow\" href=\"https:\/\/www.dataschool.io\/cloud-services-for-jupyter-notebook\/\">Jupyter Notebook felh\u0151ben.<\/a><\/noindex>.<\/li>\n<li>A nagy neur\u00e1lis h\u00e1l\u00f3zat gy\u00e1rt\u00f3k elkezdtek l\u00e9trehozni <strong>\u201emodellek \u00e1llatkertjeit\u201d<\/strong> (model zoo). Ezek seg\u00edts\u00e9g\u00e9vel gyorsan l\u00e9tezik kereskedelmi alkalmaz\u00e1s: <noindex><a rel=\"nofollow\" href=\"https:\/\/tfhub.dev\/\">TF Hub<\/a><\/noindex> a TensorFlow-hoz, <noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/open-mmlab\/mmdetection\">MMDetection<\/a><\/noindex> a PyTorch-hoz, <noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/facebookresearch\/Detectron\">Detectron<\/a><\/noindex> a Caffe2-h\u00f6z, <noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/wkentaro\/chainer-modelzoo\">chainer-modelzoo<\/a><\/noindex> a Chainer-hez \u00e9s <noindex><a rel=\"nofollow\" href=\"https:\/\/modelzoo.co\/\">egy\u00e9b<\/a><\/noindex>.<\/li>\n<li>Val\u00f3s id\u0151ben m\u0171k\u00f6d\u0151 neur\u00e1lis h\u00e1l\u00f3zatok <strong>(real-time) mobil eszk\u00f6z\u00f6k\u00f6n. 10-50 k\u00e9pkocka m\u00e1sodpercenk\u00e9nt.<\/strong> Neur\u00e1lis h\u00e1l\u00f3zatok alkalmaz\u00e1sa telefonokban (TF Lite), b\u00f6ng\u00e9sz\u0151kben (TF.js) \u00e9s<\/li>\n<li>h\u00e1ztart\u00e1si eszk\u00f6z\u00f6kben <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/19ZNz2N79u4\">(IoT,<\/a><\/noindex> internet <strong>I<\/strong>dolgok). K\u00fcl\u00f6n\u00f6sen a telefonokban, amelyek m\u00e1r \u201ehardveres\u201d szinten is t\u00e1mogatj\u00e1k a neur\u00e1lis h\u00e1l\u00f3zatokat (neuroacceler\u00e1torok). <strong>) k\u00f6z\u00f6tt. Ezut\u00e1n haszn\u00e1l egy oper\u00e1tort a jogosults\u00e1gok hozz\u00e1ad\u00e1s\u00e1ra vagy elt\u00e1vol\u00edt\u00e1s\u00e1ra az aktu\u00e1lis m\u00f3db\u00f3l, vagy abszol\u00fat jogokat \u00e1ll\u00edt be.<\/strong>f <strong>T<\/strong>k\u00fcl\u00f6n\u00f6sen azokban a telefonokban, amelyek m\u00e1r t\u00e1mogatj\u00e1k a mesters\u00e9ges intelligenci\u00e1t \"hardver\" szinten (idegprocesszorok).<\/li>\n<li>\u00abMinden eszk\u00f6z, ruhadarab \u00e9s esetleg m\u00e9g az \u00e9tel is rendelkezni fog <strong>IP-v6 c\u00edmmel<\/strong> \u00e9s kommunik\u00e1lni fog egym\u00e1ssal\u00bb \u2013 <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/GG7H8Xa4m8I?t=85\">Sebastian Thrun<\/a><\/noindex>.<\/li>\n<li>A g\u00e9pi tanul\u00e1ssal kapcsolatos publik\u00e1ci\u00f3k sz\u00e1ma kezdte <noindex><a rel=\"nofollow\" href=\"http:\/\/data-mining.philippe-fournier-viger.com\/too-many-machine-learning-papers\">t\u00fall\u00e9pni a Moore-t\u00f6rv\u00e9nyt<\/a><\/noindex> (k\u00e9t\u00e9venk\u00e9nti megdupl\u00e1z\u00f3d\u00e1s) 2015 \u00f3ta. Nyilv\u00e1nval\u00f3an sz\u00fcks\u00e9g van cikk-elemz\u0151 neur\u00e1lis h\u00e1l\u00f3zatokra.<\/li>\n<li>A k\u00f6vetkez\u0151 technol\u00f3gi\u00e1k n\u00e9pszer\u0171s\u00e9ge n\u00f6vekszik:\n<ul>\n<li><strong>PyTorch<\/strong> \u2013 a n\u00e9pszer\u0171s\u00e9g gyorsan n\u00f6vekszik, \u00e9s \u00fagy t\u0171nik, hogy m\u00e1r megel\u0151zi a TensorFlow-t.<\/li>\n<li>Automatikus hiperparam\u00e9ter-keres\u00e9s <strong>AutoML<\/strong> \u2013 a n\u00e9pszer\u0171s\u00e9g fokozatosan n\u00f6vekszik.<\/li>\n<li>A pontoss\u00e1g fokozatos cs\u00f6kken\u00e9se \u00e9s a sz\u00e1m\u00edt\u00e1si sebess\u00e9g n\u00f6veked\u00e9se: <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/rln_kZbYaWc\">hom\u00e1lyos logika<\/a><\/noindex>, algoritmusok <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/MIPkK5ZAsms\">boosting<\/a><\/noindex>, imprec\u00edz (megk\u00f6zel\u00edt\u0151) sz\u00e1m\u00edt\u00e1sok, kvant\u00e1l\u00e1s (amikor a neur\u00e1lis h\u00e1l\u00f3zat s\u00falyait eg\u00e9sz sz\u00e1mokba konvert\u00e1lj\u00e1k \u00e9s kvant\u00e1lj\u00e1k), neur\u00e1lis gyors\u00edt\u00f3k.<\/li>\n<li>Ford\u00edt\u00e1s <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/e-WB4lfg30M\">k\u00e9pek sz\u00f6vegg\u00e9 t\u00f6rt\u00e9n\u0151 ford\u00edt\u00e1s\u00e1ra<\/a><\/noindex> \u00e9s <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/rAbhypxs1qQ\">sz\u00f6vegek k\u00e9pp\u00e9 alak\u00edt\u00e1s\u00e1ra<\/a><\/noindex>.<\/li>\n<li>M\u00f3dszer kidolgoz\u00e1sa <noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/OrHLacCDZVQ\">h\u00e1romdimenzi\u00f3s objektumok a vide\u00f3n<\/a><\/noindex>, m\u00e1r val\u00f3s id\u0151ben.<\/li>\n<li>A DL-ben a l\u00e9nyeg a sok adat, de azokat \u00f6sszegy\u0171jteni \u00e9s c\u00edmk\u00e9zni nem k\u00f6nny\u0171. Ez\u00e9rt fejl\u0151dik az automatiz\u00e1lt c\u00edmk\u00e9z\u00e9s (<noindex><a rel=\"nofollow\" href=\"https:\/\/youtu.be\/NcKTn4C91Yc\">automated annotation<\/a><\/noindex>) neur\u00e1lis h\u00e1l\u00f3zatok sz\u00e1m\u00e1ra neur\u00e1lis h\u00e1l\u00f3zatok seg\u00edts\u00e9g\u00e9vel.<\/li>\n<\/ul>\n<\/li>\n<li>A neur\u00e1lis h\u00e1l\u00f3zatokkal a Sz\u00e1m\u00edt\u00e1stechnika hirtelen <strong>k\u00eds\u00e9rleti tudom\u00e1nny\u00e1 v\u00e1lt<\/strong> \u00e9s megjelent <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/post\/480348\">a reproduk\u00e1lhat\u00f3s\u00e1g v\u00e1ls\u00e1ga<\/a><\/noindex>.<\/li>\n<li>Az IT-p\u00e9nz \u00e9s a neur\u00e1lis h\u00e1l\u00f3zatok n\u00e9pszer\u0171s\u00e9ge egyszerre jelent meg, amikor a sz\u00e1m\u00edt\u00e1sok piaci \u00e9rt\u00e9kk\u00e9 v\u00e1ltak. A gazdas\u00e1g az aranyvalut\u00e1b\u00f3l <strong>arany-valuta-sz\u00e1m\u00edt\u00e1stechnika<\/strong>t\u00e1pl\u00e1l\u00e9kk\u00e1 v\u00e1lik. N\u00e9zd meg a cikkemet a <noindex><a rel=\"nofollow\" href=\"https:\/\/ru.wikipedia.org\/wiki\/%D0%AD%D0%BA%D0%BE%D0%BD%D0%BE%D1%84%D0%B8%D0%B7%D0%B8%D0%BA%D0%B0\">k\u00f3nofizik\u00e1r\u00f3l<\/a><\/noindex> \u00e9s az IT-p\u00e9nz megjelen\u00e9s\u00e9nek ok\u00e1r\u00f3l.<\/li>\n<\/ul>\n<p><\/p>\n<p>Fokozatosan megjelenik egy \u00faj <noindex><a rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/post\/481844\">ML\/DL programoz\u00e1si metodol\u00f3gia<\/a><\/noindex> (G\u00e9pi Tanul\u00e1s &amp; M\u00e9ly Tanul\u00e1s), amely a programot mint a betan\u00edtott neur\u00e1lis h\u00e1l\u00f3zati modellek \u00f6sszess\u00e9g\u00e9t k\u00e9pzeli el.<\/p>\n<p><\/p>\n<p><img decoding=\"async\" alt=\"Neur\u00e1lis h\u00e1l\u00f3zatok. Merre halad ez az eg\u00e9sz?\" src=\"\/wp-content\/uploads\/2020\/01\/17df76a613191522145abeba177c2685.png\" style=\"display:block;margin: 0 auto;\" \/><\/p>\n<p><\/p>\n<p>3. \u00e1bra \u2013 ML\/DL mint \u00faj programoz\u00e1si metodol\u00f3gia<\/p>\n<p><\/p>\n<p>Azonban nem jelent meg <strong>a \u00abneur\u00e1lis h\u00e1l\u00f3zatok elm\u00e9lete\u00bb<\/strong>, amely keret\u00e9ben lehetne rendszerszeml\u00e9letben gondolkodni \u00e9s dolgozni. Ami most elm\u00e9letk\u00e9nt van eml\u00edtve, az val\u00f3j\u00e1ban k\u00eds\u00e9rleti, heurisztikus algoritmusok.<\/p>\n<p><\/p>\n<p>Hivatkoz\u00e1sok az \u00e9n \u00e9s m\u00e1s forr\u00e1sokhoz:<\/p>\n<p><\/p>\n<ul>\n<li>Adat Tudom\u00e1ny h\u00edrlevel\u00e9nek feliratkoz\u00e1s. F\u0151k\u00e9nt k\u00e9pfeldolgoz\u00e1sr\u00f3l. Aki szeretne kapni, k\u00fcldj\u00f6n e-mailt (foobar167&lt;gaff-gaff&gt;gmail&lt;dot&gt;com). A cikkek \u00e9s vide\u00f3k linkjeit a gy\u0171jt\u00e9s sor\u00e1n k\u00fcld\u00f6m.<\/li>\n<li>\u00c1tfog\u00f3 <noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/foobar167\/articles\/blob\/master\/Machine_Learning\/courses_on_machine_learning.md\">kurzusok \u00e9s cikkek<\/a><\/noindex>, amelyeket elv\u00e9geztem \u00e9s amelyeket meg szeretn\u00e9k tanulni.<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/foobar167\/articles\/blob\/master\/Ubuntu\/13_Keras_and_TensorFlow_how-tos.md#exercises\">Kezd\u0151knek sz\u00f3l\u00f3 kurzusok \u00e9s vide\u00f3k<\/a><\/noindex>, amelyekkel \u00e9rdemes elkezdeni a neur\u00e1lis h\u00e1l\u00f3zatok tanulm\u00e1nyoz\u00e1s\u00e1t. Plusz egy bros\u00fara <noindex><a rel=\"nofollow\" href=\"https:\/\/foobar167.github.io\/page\/vvedeniye-v-mashinnoye-obucheniye-i-iskusstvennyye-neyronnyye-seti.html\">\u00abBevezet\u00e9s a g\u00e9pi tanul\u00e1sba \u00e9s a mesters\u00e9ges neur\u00e1lis h\u00e1l\u00f3zatokba\u00bb<\/a><\/noindex>.<\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/foobar167\/articles\/blob\/master\/Ubuntu\/13_Keras_and_TensorFlow_how-tos.md#tools\">Hasznos eszk\u00f6z\u00f6k<\/a><\/noindex>, ahol mindenki tal\u00e1l valami \u00e9rdekeset mag\u00e1nak.<\/li>\n<li>Nagyon hasznosnak bizonyultak <strong>tudom\u00e1nyos cikkek elemz\u00e9s\u00e9vel foglalkoz\u00f3 vide\u00f3csatorn\u00e1k<\/strong> a Data Science ter\u00fclet\u00e9n. Tal\u00e1ljatok r\u00e1juk, iratkozzatok fel \u00e9s adj\u00e1tok \u00e1t a linkeket a koll\u00e9g\u00e1itoknak, \u00e9s nekem is. P\u00e9ld\u00e1k:\n<ul>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/www.youtube.com\/user\/keeroyz\">Two Minute Papers<\/a><\/noindex><\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/www.youtube.com\/channel\/UCHB9VepY6kYvZjj0Bgxnpbw\">Henry AI Labs<\/a><\/noindex><\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/www.youtube.com\/channel\/UCZHmQk67mSJgfCCTn7xBfew\">Yannic Kilcher<\/a><\/noindex><\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/www.youtube.com\/channel\/UC5_6ZD6s8klmMu9TXEB_1IA\">CodeEmporium<\/a><\/noindex><\/li>\n<li><noindex><a rel=\"nofollow\" href=\"https:\/\/www.dlology.com\">Chengwei Zhang blogja<\/a><\/noindex> vagy <noindex><a rel=\"nofollow\" href=\"https:\/\/github.com\/Tony607\">Tony607<\/a><\/noindex> l\u00e9p\u00e9sr\u0151l l\u00e9p\u00e9sre \u00fatmutat\u00f3kkal \u00e9s ny\u00edlt forr\u00e1sk\u00f3d\u00fa projektekkel.<\/li>\n<\/ul>\n<\/li>\n<\/ul>\n<p><\/p>\n<p>K\u00f6sz\u00f6n\u00f6m a figyelmet!<\/p>\n<p>Forr\u00e1s: <a content=\"nofollow\" rel=\"nofollow\" href=\"https:\/\/habr.com\/ru\/post\/482794\/\">habr.com<\/a><\/p>","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"excerpt":{"rendered":"<p>\u0421\u0442\u0430\u0442\u044c\u044f \u0441\u043e\u0441\u0442\u043e\u0438\u0442 \u0438\u0437 \u0434\u0432\u0443\u0445 \u0447\u0430\u0441\u0442\u0435\u0439: \u041a\u0440\u0430\u0442\u043a\u043e\u0435 \u043e\u043f\u0438\u0441\u0430\u043d\u0438\u0435 \u043d\u0435\u043a\u043e\u0442\u043e\u0440\u044b\u0445 \u0430\u0440\u0445\u0438\u0442\u0435\u043a\u0442\u0443\u0440 \u0441\u0435\u0442\u0435\u0439 \u043f\u043e \u043e\u0431\u043d\u0430\u0440\u0443\u0436\u0435\u043d\u0438\u044e \u043e\u0431\u044a\u0435\u043a\u0442\u043e\u0432 \u043d\u0430 \u0438\u0437\u043e\u0431\u0440\u0430\u0436\u0435\u043d\u0438\u0438 \u0438 \u0441\u0435\u0433\u043c\u0435\u043d\u0442\u0430\u0446\u0438\u0438 \u0438\u0437\u043e\u0431\u0440\u0430\u0436\u0435\u043d\u0438\u0439 \u0441 \u0441\u0430\u043c\u044b\u043c\u0438 \u043f\u043e\u043d\u044f\u0442\u043d\u044b\u043c\u0438 \u0434\u043b\u044f \u043c\u0435\u043d\u044f \u0441\u0441\u044b\u043b\u043a\u0430\u043c\u0438 \u043d\u0430 \u0440\u0435\u0441\u0443\u0440\u0441\u044b. \u0421\u0442\u0430\u0440\u0430\u043b\u0441\u044f \u0432\u044b\u0431\u0438\u0440\u0430\u0442\u044c \u0432\u0438\u0434\u0435\u043e \u043f\u043e\u044f\u0441\u043d\u0435\u043d\u0438\u044f \u0438 \u0436\u0435\u043b\u0430\u0442\u0435\u043b\u044c\u043d\u043e \u043d\u0430 \u0440\u0443\u0441\u0441\u043a\u043e\u043c \u044f\u0437\u044b\u043a\u0435. \u0412\u0442\u043e\u0440\u0430\u044f \u0447\u0430\u0441\u0442\u044c \u0441\u043e\u0441\u0442\u043e\u0438\u0442 \u0432 \u043f\u043e\u043f\u044b\u0442\u043a\u0435 \u043e\u0441\u043e\u0437\u043d\u0430\u0442\u044c \u043d\u0430\u043f\u0440\u0430\u0432\u043b\u0435\u043d\u0438\u0435 \u0440\u0430\u0437\u0432\u0438\u0442\u0438\u044f \u0430\u0440\u0445\u0438\u0442\u0435\u043a\u0442\u0443\u0440 \u043d\u0435\u0439\u0440\u043e\u043d\u043d\u044b\u0445 \u0441\u0435\u0442\u0435\u0439. \u0418 \u0442\u0435\u0445\u043d\u043e\u043b\u043e\u0433\u0438\u0439 \u043d\u0430 \u0438\u0445 \u043e\u0441\u043d\u043e\u0432\u0435. \u0420\u0438\u0441\u0443\u043d\u043e\u043a 1 \u2013 \u041f\u043e\u043d\u0438\u043c\u0430\u0442\u044c [&hellip;]<\/p>\n","protected":false,"gt_translate_keys":[{"key":"rendered","format":"html"}]},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[702],"tags":[],"class_list":["post-54780","post","type-post","status-publish","format-standard","hentry","category-news"],"aioseo_notices":[],"aioseo_head":"\n\t\t<!-- All in One SEO 5.0.3 - aioseo.com -->\n\t<meta name=\"description\" content=\"\u0421\u0442\u0430\u0442\u044c\u044f \u0441\u043e\u0441\u0442\u043e\u0438\u0442 \u0438\u0437 \u0434\u0432\u0443\u0445 \u0447\u0430\u0441\u0442\u0435\u0439: \u041a\u0440\u0430\u0442\u043a\u043e\u0435 \u043e\u043f\u0438\u0441\u0430\u043d\u0438\u0435 \u043d\u0435\u043a\u043e\u0442\u043e\u0440\u044b\u0445 \u0430\u0440\u0445\u0438\u0442\u0435\u043a\u0442\u0443\u0440 \u0441\u0435\u0442\u0435\u0439 \u043f\u043e \u043e\u0431\u043d\u0430\u0440\u0443\u0436\u0435\u043d\u0438\u044e \u043e\u0431\u044a\u0435\u043a\u0442\u043e\u0432 \u043d\u0430 \u0438\u0437\u043e\u0431\u0440\u0430\u0436\u0435\u043d\u0438\u0438 \u0438 \u0441\u0435\u0433\u043c\u0435\u043d\u0442\u0430\u0446\u0438\u0438 \u0438\u0437\u043e\u0431\u0440\u0430\u0436\u0435\u043d\u0438\u0439 \u0441 \u0441\u0430\u043c\u044b\u043c\u0438 \u043f\u043e\u043d\u044f\u0442\u043d\u044b\u043c\u0438 \u0434\u043b\u044f \u043c\u0435\u043d\u044f \u0441\u0441\u044b\u043b\u043a\u0430\u043c\u0438 \u043d\u0430 \u0440\u0435\u0441\u0443\u0440\u0441\u044b.\" \/>\n\t<meta name=\"robots\" content=\"max-image-preview:large\" \/>\n\t<meta name=\"author\" content=\"Yuri Gagarin\"\/>\n\t<link rel=\"canonical\" href=\"https:\/\/prohoster.info\/hu\/blog\/news\/nejroseti-kuda-eto-vse-dvizhetsya\" \/>\n\t\t<meta name=\"generator\" content=\"All in One SEO (AIOSEO) 5.0.3\" \/>\n\t\t<meta property=\"og:locale\" content=\"hu_HU\" \/>\n\t\t<meta property=\"og:site_name\" content=\"ProHoster | \u041a\u0443\u043f\u0438\u0442\u044c \u043d\u0430\u0434\u0435\u0436\u043d\u044b\u0439 \u0445\u043e\u0441\u0442\u0438\u043d\u0433 \u0434\u043b\u044f \u0441\u0430\u0439\u0442\u043e\u0432 \u0441 \u0437\u0430\u0449\u0438\u0442\u043e\u0439 \u043e\u0442 DDoS, VPS VDS \u0441\u0435\u0440\u0432\u0435\u0440\u044b\" \/>\n\t\t<meta property=\"og:type\" content=\"article\" \/>\n\t\t<meta property=\"og:title\" content=\"\ud83e\udd47\u041d\u0435\u0439\u0440\u043e\u0441\u0435\u0442\u0438. \u041a\u0443\u0434\u0430 \u044d\u0442\u043e \u0432\u0441\u0435 \u0434\u0432\u0438\u0436\u0435\u0442\u0441\u044f | ProHoster\" \/>\n\t\t<meta property=\"og:description\" content=\"\u0421\u0442\u0430\u0442\u044c\u044f \u0441\u043e\u0441\u0442\u043e\u0438\u0442 \u0438\u0437 \u0434\u0432\u0443\u0445 \u0447\u0430\u0441\u0442\u0435\u0439: \u041a\u0440\u0430\u0442\u043a\u043e\u0435 \u043e\u043f\u0438\u0441\u0430\u043d\u0438\u0435 \u043d\u0435\u043a\u043e\u0442\u043e\u0440\u044b\u0445 \u0430\u0440\u0445\u0438\u0442\u0435\u043a\u0442\u0443\u0440 \u0441\u0435\u0442\u0435\u0439 \u043f\u043e \u043e\u0431\u043d\u0430\u0440\u0443\u0436\u0435\u043d\u0438\u044e \u043e\u0431\u044a\u0435\u043a\u0442\u043e\u0432 \u043d\u0430 \u0438\u0437\u043e\u0431\u0440\u0430\u0436\u0435\u043d\u0438\u0438 \u0438 \u0441\u0435\u0433\u043c\u0435\u043d\u0442\u0430\u0446\u0438\u0438 \u0438\u0437\u043e\u0431\u0440\u0430\u0436\u0435\u043d\u0438\u0439 \u0441 \u0441\u0430\u043c\u044b\u043c\u0438 \u043f\u043e\u043d\u044f\u0442\u043d\u044b\u043c\u0438 \u0434\u043b\u044f \u043c\u0435\u043d\u044f \u0441\u0441\u044b\u043b\u043a\u0430\u043c\u0438 \u043d\u0430 \u0440\u0435\u0441\u0443\u0440\u0441\u044b.\" \/>\n\t\t<meta property=\"og:url\" content=\"https:\/\/prohoster.info\/hu\/blog\/news\/nejroseti-kuda-eto-vse-dvizhetsya\" \/>\n\t\t<meta property=\"og:image\" content=\"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg\" \/>\n\t\t<meta property=\"og:image:secure_url\" content=\"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg\" \/>\n\t\t<meta property=\"og:image:width\" content=\"350\" \/>\n\t\t<meta property=\"og:image:height\" content=\"350\" \/>\n\t\t<meta property=\"article:published_time\" content=\"2020-01-03T21:00:00+00:00\" \/>\n\t\t<meta property=\"article:modified_time\" content=\"2020-02-18T11:02:50+00:00\" \/>\n\t\t<meta property=\"article:publisher\" content=\"https:\/\/www.facebook.com\/prohoster\" \/>\n\t\t<meta property=\"article:author\" content=\"https:\/\/www.facebook.com\/prohoster\" \/>\n\t\t<!-- All in One SEO -->\n\n","aioseo_head_json":{"title":"\ud83e\udd47Mesters\u00e9ges intelligencia. Hov\u00e1 tart ez az eg\u00e9sz | ProHoster","description":"A cikk k\u00e9t r\u00e9szb\u0151l \u00e1ll: N\u00e9h\u00e1ny h\u00e1l\u00f3zati architekt\u00fara r\u00f6vid le\u00edr\u00e1sa az objektumok k\u00e9pekben t\u00f6rt\u00e9n\u0151 \u00e9szlel\u00e9s\u00e9re \u00e9s a leg\u00e9rthet\u0151bb forr\u00e1sokkal val\u00f3 k\u00e9pkategoriz\u00e1l\u00e1s.","canonical_url":"https:\/\/prohoster.info\/hu\/blog\/news\/nejroseti-kuda-eto-vse-dvizhetsya","robots":"max-image-preview:large","keywords":"","webmasterTools":{"miscellaneous":""},"schema":null,"og:locale":"hu_HU","og:site_name":"ProHoster | \u041a\u0443\u043f\u0438\u0442\u044c \u043d\u0430\u0434\u0435\u0436\u043d\u044b\u0439 \u0445\u043e\u0441\u0442\u0438\u043d\u0433 \u0434\u043b\u044f \u0441\u0430\u0439\u0442\u043e\u0432 \u0441 \u0437\u0430\u0449\u0438\u0442\u043e\u0439 \u043e\u0442 DDoS, VPS VDS \u0441\u0435\u0440\u0432\u0435\u0440\u044b","og:type":"article","og:title":"\ud83e\udd47\u041d\u0435\u0439\u0440\u043e\u0441\u0435\u0442\u0438. \u041a\u0443\u0434\u0430 \u044d\u0442\u043e \u0432\u0441\u0435 \u0434\u0432\u0438\u0436\u0435\u0442\u0441\u044f | ProHoster","og:description":"\u0421\u0442\u0430\u0442\u044c\u044f \u0441\u043e\u0441\u0442\u043e\u0438\u0442 \u0438\u0437 \u0434\u0432\u0443\u0445 \u0447\u0430\u0441\u0442\u0435\u0439: \u041a\u0440\u0430\u0442\u043a\u043e\u0435 \u043e\u043f\u0438\u0441\u0430\u043d\u0438\u0435 \u043d\u0435\u043a\u043e\u0442\u043e\u0440\u044b\u0445 \u0430\u0440\u0445\u0438\u0442\u0435\u043a\u0442\u0443\u0440 \u0441\u0435\u0442\u0435\u0439 \u043f\u043e \u043e\u0431\u043d\u0430\u0440\u0443\u0436\u0435\u043d\u0438\u044e \u043e\u0431\u044a\u0435\u043a\u0442\u043e\u0432 \u043d\u0430 \u0438\u0437\u043e\u0431\u0440\u0430\u0436\u0435\u043d\u0438\u0438 \u0438 \u0441\u0435\u0433\u043c\u0435\u043d\u0442\u0430\u0446\u0438\u0438 \u0438\u0437\u043e\u0431\u0440\u0430\u0436\u0435\u043d\u0438\u0439 \u0441 \u0441\u0430\u043c\u044b\u043c\u0438 \u043f\u043e\u043d\u044f\u0442\u043d\u044b\u043c\u0438 \u0434\u043b\u044f \u043c\u0435\u043d\u044f \u0441\u0441\u044b\u043b\u043a\u0430\u043c\u0438 \u043d\u0430 \u0440\u0435\u0441\u0443\u0440\u0441\u044b.","og:url":"https:\/\/prohoster.info\/hu\/blog\/news\/nejroseti-kuda-eto-vse-dvizhetsya","og:image":"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg","og:image:secure_url":"https:\/\/prohoster.info\/wp-content\/uploads\/2021\/11\/logo-350.jpg","og:image:width":350,"og:image:height":350,"article:published_time":"2020-01-03T21:00:00+00:00","article:modified_time":"2020-02-18T11:02:50+00:00","article:publisher":"https:\/\/www.facebook.com\/prohoster","article:author":"https:\/\/www.facebook.com\/prohoster"},"aioseo_meta_data":{"post_id":"54780","title":null,"description":null,"keywords":null,"keyphrases":null,"primary_term":null,"canonical_url":null,"og_title":null,"og_description":null,"og_object_type":"default","og_image_type":"default","og_image_url":null,"og_image_width":null,"og_image_height":null,"og_image_custom_url":null,"og_image_custom_fields":null,"og_video":null,"og_custom_url":null,"og_article_section":null,"og_article_tags":null,"twitter_use_og":false,"twitter_card":"default","twitter_image_type":"default","twitter_image_url":null,"twitter_image_custom_url":null,"twitter_image_custom_fields":null,"twitter_title":null,"twitter_description":null,"schema":{"blockGraphs":[],"customGraphs":[],"default":{"data":{"Article":[],"Course":[],"Dataset":[],"FAQPage":[],"Movie":[],"Person":[],"Product":[],"ProductReview":[],"Car":[],"Recipe":[],"Service":[],"SoftwareApplication":[],"WebPage":[]},"graphName":"","isEnabled":true},"graphs":[]},"schema_type":null,"schema_type_options":null,"pillar_content":false,"robots_default":true,"robots_noindex":false,"robots_noarchive":false,"robots_nosnippet":false,"robots_nofollow":false,"robots_noimageindex":false,"robots_noodp":false,"robots_notranslate":false,"robots_max_snippet":null,"robots_max_videopreview":null,"robots_max_imagepreview":"large","priority":null,"frequency":null,"local_seo":null,"seo_analyzer_scan_date":"2026-01-24 12:43:20","breadcrumb_settings":null,"limit_modified_date":false,"reviewed_by":null,"ai":null,"created":"2021-02-28 19:57:38","updated":"2026-01-24 12:43:20","focus_keyword":null,"additional_keywords":null,"truseo_locale":null},"gt_translate_keys":[{"key":"link","format":"url"}],"_links":{"self":[{"href":"https:\/\/prohoster.info\/hu\/wp-json\/wp\/v2\/posts\/54780","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/prohoster.info\/hu\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/prohoster.info\/hu\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/prohoster.info\/hu\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/prohoster.info\/hu\/wp-json\/wp\/v2\/comments?post=54780"}],"version-history":[{"count":0,"href":"https:\/\/prohoster.info\/hu\/wp-json\/wp\/v2\/posts\/54780\/revisions"}],"wp:attachment":[{"href":"https:\/\/prohoster.info\/hu\/wp-json\/wp\/v2\/media?parent=54780"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/prohoster.info\/hu\/wp-json\/wp\/v2\/categories?post=54780"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/prohoster.info\/hu\/wp-json\/wp\/v2\/tags?post=54780"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}