Indlela umbono womshini osebenza ngayo nokuthi usetshenziselwa ini

Siphila sizungezwe izinhlelo ezikwazi ukubona, ukuqaphela, nokusabela ngokushesha njengomuntu, yize zivame ukunganakwa. Kusukela kuselula evula isikrini sayo ngobuso bakho kuya kumshini wezimboni obona izingxenye ezinephutha ngokushesha, zonke zithembele ku... ubuchwepheshe bokubona komshini obusebenzisa ubuhlakani bokwenziwa abashiye ilabhorethri ukuze babe yingxenye yokuphila kwansuku zonke.

Nakuba kungase kubonakale sengathi kuyindlela entsha yobuchwepheshe, iqiniso liwukuthi ukuhlakanipha kokwenziwa kanye nombono wekhompyutha sekukhona isikhathi eside. amashumi eminyaka ethuthuka njengezifundo zesayensiUmehluko ukuthi manje, ngenxa yamandla okusebenzisa ikhompyutha kanye nokukhula kwe- ukufunda okujulileAmandla ayo ayasetshenziswa ngempela: kungenzeka ukuqeqesha amamodeli ngaphandle kokuba unjiniyela, ukusetshenziswa kwayo kwentando yeningi ezinkampanini yanoma yiluphi usayizi, futhi ngaphezu kwakho konke, ukwenza izinqumo ezizenzakalelayo ezazincike embonweni womuntu ngaphambili.

Kuyini ngempela ukubona kwekhompyutha?

Ngobuchwepheshe, umbono wekhompyutha (noma umbono wekhompyutha) igatsha le-AI elibhekene ne- ukuthwebula, ukucubungula, ukuhlaziya nokuqonda izithombe namavidiyo kusuka ezweni langempela ukuze lihunyushwe libe idatha yezinombolo noma engokomfanekiso umshini ongayiphatha. Okusho ukuthi, liguqula amaphikseli abe ulwazi oluhlelekile: izinto, izigaba, izikhundla, ama-anomalies, amaphethini, njll.

Uma ubuhlakani bokwenziwa buhlose ukwenza izinhlelo zokubala ukucabanga nokwenza izinqumo ngokuzimelaUmbono wokwenziwa ubanika amehlo: ubavumela ukuthi bathole ulwazi olubonakalayo endaweni ezungezile, baluchaze, futhi yenza ngokufanele ngaphandle kokungenela komuntu ngqoNgale ndlela, uhlelo lunganquma, isibonelo, ukuthi i-X-ray ikhombisa yini i-pneumonia engaba khona noma ukuthi umkhiqizo osemgqeni wokuhlanganisa awukho yini esinqumweni.

Ngokwendlela engokoqobo, ukusebenzisa umbono womshini kuhilela imisebenzi yokuthola, ukuhlela, nokulandelela izithombe noma amavidiyo ngokuzenzakalela okuthi, uma kwenziwa ngumuntu oyedwa, kuzodinga isikhathi, ukunakwa okuqhubekayo, kanye nezinga eliphezulu lobuchwepheshe. Ngaphezu kwalokho, njengoba zisekelwe emithethweni yezibalo kanye nezibalo, Kunciphisa ukuzikhethela kanye nokubandlulula okukhona esweni lomuntu.Kunciphisa amaphutha futhi kusiza ekulinganiseni izindinganiso zekhwalithi noma zokuphepha.

Konke lokhu kuholela ezinzuzweni ezibonakalayo kakhulu ezinhlanganweni: Izindleko eziphansi, amaphutha ambalwa, kanye nezinqumo ezisheshayo ezisekelwe kudatha ebonakalayoFuthi, njengebhonasi, ikuvumela ukuthi usebenzise izithombe eziningi kakhulu okungenakwenzeka ukuzibuyekeza ngesandla, into ebalulekile enkathini yamanje ye-big data kanye nokuxhumana okukhulu.

Indlela umbono womshini osebenza ngayo isinyathelo ngesinyathelo

Umbono wokwenziwa uzama ukulingisa, empeleni, inqubo yokubona komuntu.Okokuqala, ithwebula indawo, bese iyiguqula ibe yizimpawu uhlelo olungazicubungula, bese ibona amaphethini, bese ekugcineni ikhiqiza impendulo. Umehluko oyinhloko ukuthi, esikhundleni sobuchopho bezinto eziphilayo, ithembele kuma-algorithms e-AI kanye namanethiwekhi ajulile e-neural.

Ukuze le nqubo isebenze, kudingeka amabhlogo amabili amakhulu: ngakolunye uhlangothi, izingxenye zomzimba zeqoqo (amakhamera, izinzwa, ukukhanyisa, ama-converter) futhi, ngakolunye uhlangothi, Amamodeli e-AI acubungula futhi aqonde isithombeZombili zisebenza ngokubambisana ukuguqula isithombe noma uhlaka lwevidiyo olulula lube ulwazi olusebenzisekayo.

Ukuthwebula idatha: amakhamera, izinzwa kanye nokwenza kube yidijithali

Isixhumanisi sokuqala kulolu chungechunge yi-hardware. Uhlelo lwesimanje lokubona komshini luhlanganisa amakhamera edijithali, izinhlelo zokukhanyisa ezilawulwayo, izinzwa namadivayisi okuthwebula uhlaka abanomthwalo wemfanelo wokuthatha izithombe zekhwalithi eyanele ukuze zihlaziywe kamuva.

Amakhamera akhiqiza isithombe se-analog sendawo, bese sidlula ku- isiguquli se-analog-to-digitalLe ngxenye iguqula ukukhanya okuqoqiwe kube i-matrix yamanani ezinombolo amelela amaphikseli kwesithombe. Iphikseli ngayinye ingabhala ulwazi lokuqina (ngomnyama nomhlophe) noma ulwazi lombala (isibonelo, ngefomethi ye-RGB).

Ezindaweni zezimboni noma ezenzakalelayo ezithuthukisiwe, kuvamile kakhulu ukuhlanganisa lesi sithombe esithathwe kanye ezinye izinhlelo zokwenza izinto ngokuzenzakalelayo kanye nokunyakaza: amarobhothi abeka izingcezu phambi kwekhamera, amabhande okuhambisa ahambisana nokukhishwa kwe-shutter yekhamera, noma izinhlelo zemishini ezilungisa ukugxila nokukhanya ukuze ziqinisekise njalo izimo ezinhle kakhulu.

Lesi sigaba sokuqala singabonakala sincane, kodwa sibalulekile: uma idatha ebonakalayo engena ohlelweni imbi, inomsindo, noma ingaguquguqukiNoma ngabe amamodeli e-AI ayinkimbinkimbi kangakanani, umphumela ngeke uthembeke. Yingakho amaphrojekthi amakhulu ombono womshini etshala umzamo omkhulu ekwakhiweni nasekulinganisweni kwezingxenye zokubona kanye nokutholwa kwedatha. Ukufakwa okuningi okulula kusebenzisa ngisho namadivayisi kanye nezisheshisi ezihambisana ne-AI. Ukusajingijolo Pi ukuze kwenziwe ama-prototyping kanye nokusetshenziswa okuncane.

Ubuchwepheshe obubalulekile: ukufunda okujulile kanye namanethiwekhi e-convolutional neural

Uma isithombe sesishintshiwe saba yidijithali, ingxenye "engabonakali" iyaqala ukusebenza: ama-algorithm. Namuhla, umbono wekhompyutha wanamuhla uncike kakhulu ku- ukufunda okujulile kanye namanethiwekhi e-convolutional neural (ama-CNN)ezisuse amasu amaningi akudala ngokusekelwe emithethweni yesandla.

Ukufunda okujulile uhlobo ukufunda komshini okusekelwe kunethiwekhi ye-neural enezingqimba eziningiNgesikhathi sokuqeqeshwa, imodeli ithola izinkulungwane noma izigidi zezithombe ezinelebula (isb., "imoto", "umhambi ngezinyawo", "ingxenye enesici", "isimila", "amaphaphu ane-pneumonia") futhi ifunda ukuqaphela amaphethini ahlukanisa isigaba esisodwa kwesinye, ngaphandle kokuthi umuntu ahlele ngesandla ukuthi imiphi imiphetho noma izimo okufanele azibheke.

Amanethiwekhi e-neural e-Convolutional aklanyelwe ngqo ukusebenza ngedatha ebonakalayo. Esikhundleni sokuphatha isithombe njengohlu oluyisicaba lwezinombolo, Basebenzisa isakhiwo samaphikseli esinobukhulu obubili bese usebenzisa izihlungi zasendaweni (izinhlayiya) ezishelela esithombeni ukuze uthole izici ezibonakalayo: imiphetho, ukuthungwa, amakhona, amaphethini aphindaphindayo, njll.

Ku-CNN ejwayelekile sithola okungenani izinhlobo ezintathu zezendlalelo: izendlalelo ezihlanganisiwe, izendlalelo zokuhlanganisa, kanye nezendlalelo ezixhunywe ngokugcweleEzokuqala zenza ukukhipha izici ngokusebenzisa izihlungi; ezesibili zinciphisa ubukhulu ngenkathi zigcina ulwazi olufanele kakhulu; kanti ezokugcina zihlanganisa konke okufundiwe ukukhiqiza umphumela, njengokwazi kweklasi.

Indlela i-CNN "ebona ngayo": ukuhlangana, amamephu ezici, kanye nokuhlanganisa

Ngokombono wezibalo, i-CNN ibheka isithombe njenge-matrix yamaphikseli bese isisebenzisa. enye i-array encane ebizwa ngokuthi isihlungi noma i-kernelLesi sihlungi sidlula esithombeni ngokubala umkhiqizo wechashazi phakathi kwamanani esihlungi namaphikseli endawo esisimbozayo endaweni ngayinye.

Ngemva kokuphothula lokhu kuhlola, imephu yokusebenzisa noma imephu yesiciLokhu kubonisa ukuthi leso sihlungi esithile siphendula ngamandla kangakanani esifundeni ngasinye sesithombe. Isihlungi ngasinye siyalungiswa, ngesikhathi sokuqeqeshwa, ukuze siphendule ngokujulile ohlotsheni oluthile lwephethini (isibonelo, imigqa evundlile, amakhona, ukuthungwa okusanhlamvu, ukuguqulwa okubushelelezi kokuqina, njll.).

Ngokubeka izingqimba eziningi ze-convolutional, inethiwekhi iyahamba ukwakha uhlu lwezici ezibonakalayo eziyinkimbinkimbi ngokwengeziweEzingqimbeni zokuqala ithola imiphetho elula, ezingqimbeni eziphakathi nendawo kanye nezingxenye, kanti ezingqimbeni ezijulile ingakwazi ukubona izinto eziphelele noma izingxenye ezithile kakhulu (njengeso, isondo noma uhlaka lwamaphaphu olusolisayo ku-X-ray).

Ukulandela lezi zendlalelo ezihlanganisiwe kuvame ukuza izendlalelo zokuqoqana noma ukuhlanganisa. Umsebenzi wawo ngu nciphisa usayizi wamamephu wesici Isibonelo, uma sithatha inani eliphakeme noma elimaphakathi ngaphakathi kwamabhulokhi amancane amaphikseli. Lokhu kucindezela ulwazi, kwenza imodeli isebenze kahle kakhulu, futhi kunikeza ukuguquguquka okuthile ekuhumusheni okuncane noma ekuguqulweni kwesithombe.

Ukusabalala kwangaphambili, umsebenzi wokulahlekelwa, kanye nokusabalala emuva

Yonke inqubo kusukela esithombeni sokufaka kuya kokukhiphayo kwemodeli yaziwa ngokuthi iphasi eliya phambiliKulesi sigaba, inethiwekhi isebenzisa ngokulandelana i-convolutions, i-activation engeyona i-linear, imisebenzi yokuhlanganisa kanye, ekugcineni, izendlalelo ezixhunywe ngokugcwele ezenza ingxenye yokuhlukaniswa noma yokubuyela emuva.

Ekupheleni kokusakazwa kwangaphambili, imodeli ikhiqiza umphumela: ekuhlukanisweni kwesithombe, lokhu kuvame ukuba yi-vector ye amathuba ahlobene nekilasi ngalinye elingenzeka (isibonelo, “okuvamile” noma “i-pneumonia” ku-x-ray yesifuba). Ukuze kuhlolwe ukuthi imodeli isebenze kahle yini, lesi sibikezelo siqhathaniswa nelebula langempela kusetshenziswa umsebenzi wokulahlekelwa lokho kulinganisa iphutha.

Inqubo yokuqeqesha ihilela ukuphindaphinda le nqubo izikhathi eziningi nokulungisa amapharamitha emodeli ukuze umsebenzi wokulahlekelwa wehle. Lokhu kwenziwa kusetshenziswa indlela eyaziwayo ye... ukusabalalisa emuvaLokhu kubala i-gradient yokulahlekelwa maqondana nesisindo ngasinye kunethiwekhi. Kusetshenziswa i-algorithm yokwenza ngcono, njengokwehla kwe-gradient, izisindo zibuyekezwa ngendlela enciphisa iphutha.

Njengoba inikezwe isikhathi kanye nedatha yokuqeqeshwa eyanele ebhalwe kahle, i-CNN ifunda ukwenza kanjalo hlukanisa amaphethini abonakalayo angabonakali kakhuluNgokwesibonelo, ekufomeni kwezokwelapha, kungabona amaphaphu angalingani, izindawo ezikhanyayo eziveza ukuvuvukala noma ukuba khona kwezindawo ezimanzi, ezinamafu noma ezingacacile, kanye nokuthungwa okungajwayelekile ngezinye izikhathi okungabonwa yiso lomuntu, okusiza ekutholakaleni kwezifo kusenesikhathi.

Kusukela ekuqashelweni okuyisisekelo kuya emisebenzini yombono womshini othuthukisiwe

Umbono wekhompyutha awugcini nje ngokusho ukuthi "okusesithombeni." Ngokwakhela phezu kwezisekelo ezifanayo nezama-CNN nokufunda okujulile, uthuthukisiwe. imisebenzi ehlukahlukene ekhethekile exazulula izinkinga ezithile emikhakheni eyahlukahlukene kakhulu.

Umsebenzi olula kakhulu yilowo ukuhlukaniswa kwezithombeIlebula elilodwa linikezwa isithombe sonke (ikati, inja, isikulufu esifanele, isikulufu esinephutha, njll.). Isinyathelo esisodwa esilandelayo yilesi ukutholwa kwentolapho, ngaphezu kokuhlonza ikilasi, into ngayinye itholakala ngaphakathi kwesithombe ngokudweba amabhokisi ahambisanayo.

Uma kudingeka ukunemba okuphezulu kwezinga lephikseli, okulandelayo kusetshenziswa: ukuhlukaniswa kwesibonelookukhiqiza imaski yento ngayinye, noma ngabe bangabesigaba esifanayo. Leli khono libalulekile, isibonelo, ku ukuhlaziywa kwesithombe sezokwelaphalapho kubalulekile ukuhlukanisa nokulinganisa amathumba, izicubu noma izitho ngokunembile.

Omunye umsebenzi obanzi kakhulu yi- ukulinganiselwa kokumaLobu buchwepheshe bubona amaphuzu abalulekile (amalunga, izitho zomzimba, njll.) emizimbeni yabantu noma kwezinye izinto ezihlanganisiwe. Busetshenziswa kwezemidlalo, e-ergonomics, e-augmented reality, kanye nezinhlelo zokuphepha eziqapha ukuma kwezisebenzi ukuze kuvinjelwe ukulimala noma izingozi.

Umbono wekhompyutha, ukufunda komshini nokufunda okujulile: ukuthi zihluke kanjani

Izingxoxo eziningi zihlanganisa imiqondo efana ubuhlakani bokwenziwa, ukufunda komshini kanye nokufunda okujulile sengathi zingamagama afanayo, okudala ukudideka okukhulu. Ukuqonda ubudlelwano bazo kusiza ekubekeni umbono wekhompyutha ngendlela efanele ngaphakathi kwale ndawo.

Ubuhlakani bokwenziwa yigama elibanzi kakhulu: lihlanganisa noma yiluphi ubuchwepheshe obuvumela umshini ukuthi... ukwenza imisebenzi esiyihlanganisa nokuhlakanipha kwabantu (ukucabanga, ukufunda, ukuhlela, ukuhumusha ulimi, ukubona, njll.). Ngaphakathi kwale nsimu, ukufunda komshini kuyisethi yezindlela ezivumela uhlelo ukuthi... Funda kudatha ngaphandle kokuhlela ngokucacile ngemithetho eqinile.

Ukufunda komshini kuhlanganisa ama-algorithm amaningi (izihlahla zezinqumo, imishini yokusekela ama-vector, ama-regression, njll.) angasetshenziswa ezinkingeni eziningi ezahlukene: ukubikezela ingozi ezenzakalelayo, ukuhlukanisa ama-imeyili njengogaxekile noma cha, ukuncoma imikhiqizo, njll. Embonweni wekhompyutha, lezi zindlela zendabuko zisetshenziswe emisebenzini elula noma lapho inani ledatha lingelikhulu kakhulu.

Ukufunda okujulile kuyingxenye yokufunda komshini ebonakala ngokusetshenziswa kwayo amanethiwekhi amakhulu, anezingqimba eziningi ze-neuralLawa manethiwekhi anamandla kakhulu uma esebenza idatha eningi, ikakhulukazi izithombe, njengoba bekwazi ukukhipha izici ezifanele ngokwabo ngaphandle kokungenela komuntu ngokuqondile.

Ekuboneni kwekhompyutha yesimanje, ukufunda okujulile ngokuvamile kuyindlela ekhethwayo: Ivumela izinga eliphezulu kakhulu lemininingwane, ukujwayela, kanye nokuqina. uma kuqhathaniswa nezindlela zakudala, uma nje kukhona idatha eyanele kanye namandla okusebenzisa ikhompyutha. Ngokwezinga elikhulu, yiwona amandla ashukumisayo ngemuva kwentuthuko ephawulekayo embonweni wekhompyutha eminyakeni eyishumi edlule.

Umbono womshini vs. ukucutshungulwa kwesithombe

Nakuba zihlobene kakhulu, kubalulekile ukuhlukanisa phakathi ukucutshungulwa kwesithombe kanye nokubona ngekhompyuthaNgezinye izikhathi zisetshenziswa ngokushintshana, kodwa azifani. Zivame ukusebenza ndawonye, ​​kodwa ziphishekela imigomo ehlukene.

Ukucubungula izithombe kugxile ku ukuphatha isithombe kanjalo: thuthukisa ukugqama, lungisa ukukhanya, nciphisa umsindo, faka izihlungi, shintsha usayizi, njll. Umphumela walezi zinhlobo zokusebenza uvame ukuba esinye isithombe esiguquliweYilokhu okwenziwa amathuluzi amaningi okuhlela izithombe, kodwa futhi kuyisisekelo sokulungiselela izithombe ngaphambi kokuzidlulisela kumodeli ye-AI.

Ngakolunye uhlangothi, umbono wekhompyutha uthatha isithombe noma ividiyo njengokufaka bese ukhiqiza ulwazi mayelana nokuqukethwe kwaloYiziphi izinto ezivelayo, ukuthi zikuphi, uhlobo luni lwesigcawu, uma kukhona okungahambi kahle, bangaki abantu abawela umnyango, njll. Umphumela awuseyona nje enye isithombe, kodwa idatha ehlelekile noma izinqumo ezenzakalelayo.

Empeleni, izinhlelo zesimanje zokubona komshini zivame ukufaka isigaba sokucubungula isithombe okokuqala (ukwenza ukukhanya kube ngokwejwayelekile, ukunquma izindawo ezithakazelisayo, ukuhlanekezela okulungile, njll.) okusiza umsebenzi olandelayo wamanethiwekhi ajulile e-neural abhekene nokuhumusha.

Ukusetshenziswa kwangempela kombono womshini emikhakheni ehlukene

Ukuguquguquka kombono womshini kusho ukuthi izinhlelo zokusebenza zawo zifinyelela cishe kunoma iyiphi insimu lapho kunezithombe noma amavidiyo okufanele uwahlaziyeKusukela ekukhiqizweni kwezimboni kuya kwezokwelapha, okuhlanganisa ukuthengisa, amabhange, ezokuthutha, ezolimo, kanye nomkhakha womphakathi, umthelela walo ukhula unyaka nonyaka.

Izinkampani eziningi azisazibuza ukuthi zisebenzise yini umbono womshini, kodwa indlela yokukuhlanganisa ngendlela efanele ukuthuthukisa izinqubo zabo, ukunciphisa izindleko, ukwandisa ukuphepha, noma ukuqonda kangcono ukuziphatha kwamakhasimende abo. Ngezansi kunezinye zezimo zokusetshenziswa ezimele kakhulu.

Ukukhiqiza, imboni kanye nokulawulwa kwekhwalithi

Embonini yokukhiqiza, umbono womshini usuphenduke ithuluzi eliyisihluthulelo sokwenza izinto ngokuzenzakalela nokulawula ikhwalithiAmakhamera afakwe emigqeni yokukhiqiza ahlala eqapha izingxenye ezidlulayo futhi athole amaphutha ezingxenyaneni zomzuzwana.

Lezi zixazululo ziyakuvumela qapha izindawo zokusebenza ezenzakalelayo, yenza izibalo zomzimba kanye nezinhlu, kala amapharamitha ekhwalithi (ukuqeda, ubukhulu, umbala), thola izinsalela noma ukungcola bese uqinisekisa ukuthi umkhiqizo ngamunye uhlangabezana ncamashi nezidingo.

Uma kuhlanganiswa nezinye izindlela zobuchwepheshe ezifana nokunyathelisa kwe-3D noma imishini ye-CNC, umbono womshini uyasiza ukuphinda futhi ukhiqize izingxenye eziyinkimbinkimbi kakhulu ngokunemba okukhuluNgaphezu kwalokho, ngokuhlanganiswa nezinzwa ze-IoT, kusiza ukubikezela izinkinga zokulungisa, ukuhlonza ukungalingani ekusebenzeni komshini, nokuvimbela ukungasebenzi okungalindelekile.

Akugcini nje ngokuthola amaphutha omkhiqizo: futhi kungenza qapha ukusetshenziswa okufanele kwemishini yokuzivikela, thola izimo eziyingozi ezitshalweni zokukhiqiza futhi ukhiphe izixwayiso kusenesikhathi ukuze kuvinjelwe izingozi zasemsebenzini.

Ukuthengisa, ukumaketha kanye nolwazi lwamakhasimende

Ezimpahleni zokuthengisa kanye nezimpahla zabathengi, umbono womshini usetshenziswa qapha ngokucophelela imisebenzi yamakhasimende esitolo: indlela abahamba ngayo, yiziphi izindawo abazivakashelayo, ukuthi bama isikhathi esingakanani phambi kweshalofu, noma ukuthi imiphi imikhiqizo abayibukayo ngaphambi kokwenza isinqumo.

Lolu lwazi, olungaziwa futhi olucutshungulwa ngendlela ehlanganisiwe, luvumela Lungiselela ukusatshalaliswa komkhiqizo, uhlele kabusha isakhiwo sesitolo, futhi ulungise imikhankaso yokumaketha ngesilinganiso semininingwane esingenakwenzeka ukusifinyelela kusetshenziswa ukuhlaziya kwewebhu noma izinhlolovo kuphela.

Izinhlelo nazo ziyandiswa ukuzihlola kusizwa umbono wokwenziwaLezi zinhlelo zingabona izinto ngaphandle kwesidingo sokuskena amabhakhodi ngayinye ngayinye. Lokhu kuthuthukisa ulwazi lwamakhasimende, kunciphisa imigqa, futhi kuvula indlela yamamodeli ezitolo angenamali.

Ngale kwendawo yokuthengisa ebonakalayo, izinkampani zisebenzisa umbono womshini ukuze Hlaziya izithombe ezinkundleni zokuxhumana, thola izitayela ezibonakalayo, funda ukuthi imikhiqizo yabo isetshenziswa kanjani ezweni langempela bese ngaleyo ndlela ulungise isu labo lomkhiqizo noma lokuxhumana.

Ukuphepha, ukuqapha kanye nomkhakha womphakathi

Umbono womshini uyinsika eyisisekelo izinhlelo zokuphepha nezokuvikela zezikhungoAmakhamera ahlakaniphile nezinzwa ezisatshalalisiwe ziqapha izindawo zomphakathi, izindawo zezimboni ezibalulekile noma izindawo ezivinjelwe futhi zikhipha izexwayiso ezizenzakalelayo uma zibona ukuziphatha okungajwayelekile.

Lezi zinhlelo zingabona Ukuba khona kwabantu abangagunyaziwe, ukufinyelela ngaphandle kwamahora okuvula, izinto ezilahliwe, noma amaphethini aphakamisa ukuthi kungenzeka kube nesigamekoKwezinye izimo, zihlanganisa ukuqashelwa kobuso ukuze kuqinisekiswe abasebenzi noma kulawulwe ukufinyelela okuphephile.

Emkhakheni wasekhaya, umbono wekhompyutha usetshenziswa kumakhamera axhunyiwe Bayaqaphela abantu, izilwane ezifuywayo, amaphakheji alethiwe, noma ukunyakaza okungajwayelekile.ukuthumela izaziso efonini yeselula yomsebenzisi. Emsebenzini, kusiza ukuqinisekisa ukuthi abasebenzi basebenzisa imishini yokuzivikela edingekayo noma balandela imithethonqubo yokuphepha ebalulekile.

Ohulumeni kanye namadolobha ahlakaniphile bayayisebenzisa ukuze qapha ithrafikhi, lungisa izibani zethrafikhi ngendlela enamandla, thola ukwephulwa kwemithetho futhi kuthuthukiswe ukuphepha komphakathi. Iphinde ifakwe ezinhlelweni zamasiko ukuze kusebenze ngokuzenzakalelayo ukuhlolwa okubonakalayo.

Ukunakekelwa kwempilo, ukuxilongwa kanye nokuhlaziywa kwezithombe zezokwelapha

Ezokwelapha zingenye yezindawo lapho umbono wokwenziwa ukhiqiza khona ushintsho olukhulu kakhulu ekusebenzeni kwezokwelaphaAmasu okuhlaziya izithombe zezokwelapha avumela ukubonakala kwezitho nezicubu ngokunemba okukhulu futhi anikeze ukwesekwa okuqondile kochwepheshe.

Phakathi kokusetshenziswa okuvame kakhulu yi- ukutholwa kwesimila ngokuhlaziywa kwama-moles kanye nezilonda zesikhumba, lo Ukuhumusha okuzenzakalelayo kwama-x-ray (isibonelo, ukuhlonza i-pneumonia noma ukwaphuka kwamathambo) kanye nokutholakala kwamaphethini amancane ekucabangeni kwe-magnetic resonance noma ekuskeni kwe-computed tomography.

Izinhlelo ezihlonyiswe ngombono ohlakaniphile ziyasiza nciphisa izikhathi zokuxilonga, uthuthukise ukunemba futhi ubeke phambili amacala aphuthumayoZingaxhunyaniswa nezizindalwazi ezinkulu zamarekhodi ezokwelapha ukuze ziphakamise ukuxilongwa noma ukwelashwa okuhlukile okungenzeka.

Ngaphezu kwalokho, umbono womshini usetshenziswa ku amadivayisi okusiza abantu abanenkinga yokubonaokwazi ukufunda imibhalo nokuyiguqula ibe yinkulumo ngokusebenzisa ukuqashelwa kwezinhlamvu ezibonakalayo (i-OCR), noma ukuchaza indawo ngendlela elula.

Izimoto ezizimele kanye nezokuthutha

Emkhakheni wezimoto, umbono womshini uwubuchwepheshe obuyinhloko ngokuphelele izimoto ezisizayo zokushayela kanye nezimoto ezizimeleAmakhamera amaningi afakwe emotweni athwebula imvelo ngesikhathi sangempela futhi ondla amamodeli e-AI ahlala echaza lokho.

Lezi zinhlelo ziyakwazi thola abahamba ngezinyawo, ezinye izimoto, izimpawu zomgwaqo, izimpawu zomgwaqo kanye nezithiyoukukhiqiza ukumelwa kwemvelo kwe-3D ngokuhlanganisa ulwazi oluvela kumakhamera nezinye izinzwa ezifana ne-LiDAR noma i-radar.

Ezimotweni ezizimele kancane, umbono womshini nawo usetshenziselwa qapha isimo somshayeliUkuhlaziya isikhundla sekhanda, ukunyakaza komzimba ongaphezulu, kanye nokuqondisa amehlo ukuze kutholakale izimpawu zokukhathala, ukuphazamiseka, noma ukozela.

Uma kutholakala amaphethini engozi, uhlelo lungakwazi khipha izexwayiso ezizwakalayo noma ezibonwayo, vula ukudlidliza esiteringini, noma ulawule ingxenye ethile ukunciphisa isivinini nokunciphisa ingozi. Lokhu kuye kwabonakala kusebenza kahle kakhulu ekunciphiseni izingozi ezibangelwa ukukhathala.

Ezolimo kanye nomkhakha wokudla kwezolimo

Umkhakha wezolimo uthole umbono womshini njengomsizi obalulekile wokuqhubekela phambili amamodeli ezolimo ezinembile nezihlakaniphileIzithombe ezithathwe ngamasathelayithi noma ama-drone zivumela ukuhlaziywa kwezindawo ezinkulu zomhlaba ngezinga lemininingwane engacabangeki eminyakeni embalwa edlule.

Ngala mathuluzi kungenzeka qapha isimo sezitshalo, thola izifo kusenesikhathi, futhi ulawule umswakama wenhlabathi futhi ulinganisele isivuno sezitshalo kusenesikhathi. Konke lokhu kwenza kube lula ukuphathwa kahle kwezinsizakusebenza ezifana namanzi, umanyolo, kanye nezibulala-zinambuzane.

Umbono womshini nawo ufakiwe ezinhlelweni ezisebenzisa Baqapha ukuziphatha kwemfuyo.Bahlonza izilwane ezigulayo, bathole ukuzalwa, futhi balawule ukufinyelela ezindaweni ezithile. Lokhu kuzenzakalela kuthuthukisa inhlalakahle yezilwane futhi kuthuthukisa umkhiqizo ophelele wamapulazi.

Embonini yokudla, ibilokhu isetshenziswa amashumi eminyaka ukuze ikhwalithi yokulawula emigqeni yokukhiqizaHlola ukubukeka kwezithelo nemifino, buyekeza ukupakishwa bese uqinisekisa ukuphepha kokudla.

Ibhange, umshwalense kanye nokuxhumana ngocingo

Emkhakheni wezezimali, umbono womshini usetshenziswa thola izimpawu ezibonakalayo zokukhwabanisa noma ukuziphatha okungajwayelekileLokhu kusebenza kokubili emahhovisi angokoqobo kanye nasekuthengiselaneni okude. Isibonelo, isithombe sangempela somsebenzisi singaqhathaniswa nesithombe esigcinwe kumadokhumenti akhe.

Iphinde ihlanganiswe ku izinqubo zokubhalisa umshwalenselapho ukuhlolwa komonakalo ezimotweni noma ezakhiweni kungenziwa ngokuzenzakalelayo ngokwengxenye kusuka ezithombeni ezithunyelwe yiklayenti, kunciphisa isikhathi nezindleko.

Kwezokuxhumana, izinkampani zisebenzisa umbono womshini ukuze bikezela futhi uthole ukuphazamiseka kwamakhasimende ngokuhlanganisa ulwazi olubonakalayo (isb., ukusetshenziswa kwamadivayisi noma izikhungo ezithile) nolunye ulwazi lokuziphatha, okusivumela ukuthi silindele izidingo ngezipesheli kanye nokuthuthukiswa kwesevisi.

Ngaphezu kwalokho, ukuqinisekiswa ngokusebenzisa ukuqashelwa kobuso Sekusabalala njengendlela yokufinyelela ngokuphephile ezinsizakalweni zasebhange nezezinkampani, njalo kuhambisana nezinye izindlela zokuphepha.

Ukuthutha impahla, ukuthuthwa kwempahla kanye nezindlu

Kwezokuthutha, umbono womshini uyasiza qapha futhi ulandelele izimpahla ngesikhathi sangempela Asikho isidingo sokuskena ngesandla okunamandla. Amakhamera abekwe ngendlela efanele yiwo kuphela adingekayo ukuze kufundwe amalebula, kuhlonzwe amaphakheji, noma kuqinisekiswe ukuthi konke kubekwe kahle.

Ngokuhlanganisa nobuchwepheshe obufana ne-RFID, lezi zinhlelo zivumela qapha izinqolobane, phatha izindawo zokugcina impahla, futhi uthuthukise imizila yokulethwa Zisebenza kahle kakhulu. Ziwusizo futhi ekutholeni umonakalo emaphaketheni ngesikhathi sokuthuthwa.

Emkhakheni wezindlu, umbono womshini usetshenziswa ku dala izinkambo ezingokoqobo nezisebenzisanayo zamakhaya, qaphela futhi ulebule amakamelo, ulinganise izikhala bese unikeza umsebenzisi ulwazi oluningiliziwe mayelana nezici zendawo ngaphandle kwesidingo sokuvakasha okuningana ngokomzimba.

Lokhu kuhlanganiswa kwezithombe ezisezingeni eliphezulu kanye nokuhlaziywa okuhlakaniphile kusindisa isikhathi kuma-ejensi kanye nabathengi abangaba khona noma abaqashi, futhi kusiza ukuvala amadili ngokushesha okukhulu.

Imfundo, imibukiso yezohwebo, kanye nezicelo zomuntu siqu

Kwezemfundo, umbono wekhompyutha usetshenziswa linganisa izindawo ezisebenzayo, amalabhorethri abonakalayo kanye namacala angokoqobo ezivumela abafundi ukuthi babhekane nezimo eziseduze nomhlaba wobungcweti ngaphandle kokuphuma ekilasini.

Emibuthanweni yezohwebo kanye nasezingqungqutheleni, amakhamera anombono wokwenziwa avumela Hlaziya ukuziphatha kwababekhona: ukugeleza kwabantu, izindawo ezishisayo, ukuxhumana nezitendi futhi, kwezinye izimo, ngisho nokulinganisa ukusabela okuvamile ngokomzwelo kokuhlangenwe nakho okuthile.

Ezingeni lomuntu siqu, ngaphezu kwezinhlelo ezishiwo ngenhla zokusiza ukuhumusha okubonakalayo okungaboni nokusheshayo (njengalapho ukhomba ifoni yakho ephathekayo ophawini ngolunye ulimi), izinkomba zokubona okwenziwe izinhlelo zokusebenza ze-augmented reality, izihlungi zemidiya yezenhlalo, kanye nemidlalo exhumanayo lokho kuncike ekuqondeni ngesikhathi sangempela ukuthi yini ephambi kwekhamera.

Konke lokhu kubonisa ukuthi ukubona ngekhompyutha akuyona into eyenzeka elabhorethri, kodwa ubuchwepheshe obuhlanganisa yonke indawo obunomthelela oqondile emnothweni, ekuphepheni, nasempilweni yansuku zonkeamandla akhe esisanda kuqala ukuwasebenzisa.

Sekukonke, umbono wekhompyutha uhlanganisa izinzwa, amakhamera, kanye neziguquli ezine-algorithms yokufunda okujulile kanye namanethiwekhi e-convolutional neural ukuze guqula izithombe namavidiyo abe ulwazi oluwusizoUkushintsha izinqumo ngokuzenzakalelayo nokwandisa ukunemba kanye nesivinini sezinqubo emikhakheni eyahlukahlukene kakhulu. Ikhono layo lokufunda kusuka kudatha enkulu ebonakalayo, ukunciphisa ukuthambekela komuntu, kanye nokuthola amaphethini angabonakali ngeso kuyenza ibe yingxenye ebalulekile ye uhlelo lwe-ecosystem lobuhlakani bokwenziwa yesimanje futhi njengethuluzi elibalulekile lezinkampani nezinhlangano lokuthola ukuncintisana, ukuthuthukisa ukuphepha kanye nokuhlinzeka ngezinsizakalo ezisebenza kahle futhi ezenzelwe wena.