Shabakadaha Transformer Space
Shabakadaha Transformer Networks (STNs) waa qaybo la baran karo oo u oggolaanaya shabakada neural inay si firfircoon u rogrogto, beddesho, dalagga, ama dib u cabbirto taladeeda si ay diiradda u saarto waxa muhiimka ah.
Dulmar
They give CNNs a built-in sense of spatial attention and invariance.
quusid qoto dheer
Shabakadda is-daba-joogga ah ee caadiga ah ayaa kaliya si liidata u beddelan isbeddellada booska, cabbirka, iyo wareegga, iyaga oo ku tiirsan isu geynta dulqaad yar. Shabakadaha Transformer Spatial, oo ay soo bandhigtay Jaderberg et al. sanadka 2015, hagaajin tan adigoo gelinaya qayb ka duwanaan kara kaas oo sameeya isbeddel joomatari oo cad oo ku saabsan khariidadaha muuqaalka. Cutubku waxa uu leeyahay saddex qaybood: shabakad deegaan oo saadaalisa halbeegyada isbeddelka, koronto-dhaliye ka dhisa shabagga muunad ee cabbirradaas, iyo muunad ka dhexgeliya gelinta dhibcaha xajmiga. Sababtoo ah tillaabo kasta waa la kala duwanaan karaa, transformer-ka oo dhan waxaa lagu tababaraa dhamaadka-ilaa-dhamaadka iyada oo dib loo faafinayo iyada oo aan lahayn kormeer dheeraad ah. Shabakadu waxa ay barataa, tusaale ahaan, in ay toosiso tirooyinka leexleexan ama ay soo dhaweeyaan gobolka ay khusayso, iyada oo kor u qaadaysa saxnaanta iyo adkaynta.
Aragtida Farsamada
Shabakadda deegaanku waxay soo saartaa cabbirro (badanaa 2x3 affine matrix) ee tarjumaadda, cabbirka, wareegga, iyo xiirida. Koronto dhaliyaha xadhkaha goostay waxa uu dib ugu celinayaa isha isku xidhka shaxdaas. Muunadle-qaaduhu wuxuu markaa akhriyaa gelinta isagoo isticmaalaya is-dhex-galka laba-geesoodka ah, kaas oo la kala sooci karo si gradients-ku u qulqulo shabakadda meelaynta. Tani waxay u oggolaanaysaa cutubku inuu barto isbeddellada oo keliya luminta hawsha, ka qaybgalka iyo kala-soocidda gobollada khuseeya.
Saamaynta Istiraatijiyadeed
Xawaaraha iyo miisaanka
Visual AI wuxuu si otomaatig ah u samayn karaa baadhista, ogaanshaha, iyo sumadaynta hawlaha miisaanka.
Xulashada dhismayaasha
Kooxaha hal-abuurka leh waxay hindise karaan fikradaha si dhakhso leh iyagoo leh dib-u-eegis buugeed yar.
Kooxda iyo socodka shaqada
Hawlgalladu waxay isticmaali karaan calaamadaha muuqaalka iyo muuqaalka kuwaas oo markii hore adkeyd in la farsameeyo.
Mustaqbalka Shabakadaha Transformer Space
STNs ayaa saameeyay sida shabakadaha ay u maamulaan joomatari iyo feejignaanta, iyaga oo u quudinaya isbedelo qallafsan iyo qaybo la bartay. In kasta oo beddelayaasha is-fiirsashada ay hadda xukumaan, muunad kala duwanaansho ee qaabka STN ayaa ku sii socota hawlaha u baahan toosinta joomatari ee cad: aqoonsiga qoraalka, kala soocida wanaagsan, iyo keenista caadi. Filo dagaal lagu kala duwanaan karo si uu uga soo muuqdo aragtida 3D, samaynta neerfaha, iyo diiwaangelinta sawirka caafimaadka, oo inta badan lagu daro feejignaan halkii lagu beddeli lahaa.
Dhaqangelinta Adduunka-dhabta ah
Toosin iyo toosinta qoraalka qaloocan ama rogrogan ka hor inta aan la aqoonsan goobta-qoraalka nidaamyada OCR
Ku soo dhowaynta gobollada takoorka ah (sida garka shimbiraha ama garabka) ee kala soocidda sawirka
Caadiyan toosinta wejiga iyo toosinta sidii tallaabo horudhac u ah dhuumaha aqoonsiga wejiga
Sixitaanka qallooca iyo toosinta baarista ee diiwaangelinta sawirka caafimaadka
Khatarta & Dariiqyada Ilaalada
Xuquuqda sawirka iyo ogolaanshaha waxay noqon kartaa khataro sharci ah haddii caddayntu aanay caddayn.
Waxqabadka moodeelku wuu ku kala duwanaan karaa iftiinka, tirakoobka, iyo deegaanka.
Wanaagga beenta ah waxa laga yaabaa inaan la dareemin ilaa xadka kalsoonida aan la kormeerin.
Qorshe Hawleedka Dhaqangelinta
Qeex shuruudaha aqbalida ee saxnaanta, dib u celinta, iyo kharashyada khaladka.
Ku tijaabi xogta ku habboon xaaladaha wax soo saarka dhabta ah.
Ku dar dib u eegis bini'aadamka si aad u hesho kalsoonida hoose ama saameeynta sare.
Lasoco moodeel dhaqaaqa oo dib u cusboonaysii kamarada ama xogta kaydinta ka dib.
Sii wad Sahaminta
Free newsletter
Get the daily AI briefing
Three verified AI stories every weekday morning, written in plain English. Free forever, no ads.
One email each weekday. Unsubscribe in one click. We never sell or share your address.
Test yourself
Take the Spatial Transformer Networks quiz
Instant feedback on every answer, and a shareable certificate with a verifiable ID once you pass a course.
Support free AI education. AI Understanding is a 501(c)(3) nonprofit — no ads, no paywall, ever. Make a donation
Hagaha xiga
DETR Ogaanshaha Transformer
Su'aalaha soo noqnoqda
What is Spatial Transformer Networks?
Shabakadaha Transformer Networks (STNs) waa qaybo la baran karo oo u oggolaanaya shabakada neural inay si firfircoon u rogrogto, beddesho, dalagga, ama dib u cabbirto taladeeda si ay diiradda u saarto waxa muhiimka ah. Waxay CNN-yada siinayaan dareen ku dhisan dareenka goobta iyo is-beddelka.
Awood noocee ah ayay Shabakadda Transformer Network ku kordhisaa shabakada neerfaha?
STNs waxay geliyaan cutub la baran karo oo tarjumi kara, rogi kara, cabbiri kara, ama khariidado muujin kara si ay diiradda u saaraan nuxurka khuseeya.
Waa kuwee saddexda qaybood ee ka kooban moduleka beddelka?
STN waxa uu ka kooban yahay shabakad deegaan (wuxuu saadaaliyaa halbeegyada), koronto-dhaliye (dhisaya isku-duwayaasha muunada), iyo muunad-qaade (isku xidha gelinta).
Maxay tahay sababta Shabakadda Transformer-ka loo tababari karaa faafinta caadiga ah?
Isdhexgalka laba-geesoodka ah ee muunadiyaha waa la kala duwanaan karaa, sidaa darteed gradients-ku waxay dib ugu qulqulaan koronto-dhaliye ilaa shabakad deegaan, taasoo awood u siinaysa tababarka dhamaadka-ilaa-dhamaadka.
Maxay shabakada deegaanku sida caadiga ah u soo saartaa isbeddelka affine?
Isbeddellada affine, shabakada deegaanku waxay saadaalisaa lix qiime oo loo habeeyey sidii 2x3 matrix codeing tarjumaad, miisaan, wareeg, iyo xiirid.
Waa maxay sababta CNN-yada caadiga ah ay si daciifnimo ah ugu beddelan yihiin isbeddellada goobta, dhiirigelinta STNs?
CNN-yadu waxay ku gaadhaan ismaandhaaf dhex-dhexaad ah oo keliya iyada oo la isu geeyo; STNs waxay ku daraan si cad, hab la baran karo oo lagu maareeyo booska, cabbirka, iyo wareejinta.