Izicelo Ezi-5 Ezivame Kakhulu Zombono Womshini

I. Isingeniso kuMbono Womshini

Umbono womshini, itshe eliyisisekelo lobuhlakani bokwenziwa, unika amandla amakhompyutha ukuhumusha idatha ebonakalayo ngokunemba okufana nokwabantu. Ngokungafani nokucubungula izithombe okuvamile, isebenzisa ukufunda okujulile kanye namanethiwekhi ezinzwa ukuhlaziya izithombe namavidiyo, iguqula izimboni ngokuzenzakalelayo imisebenzi eyinkimbinkimbi. Kusukela ekuboneni amaphutha emigqeni yokuhlangana kuya ekuxilongeni izifo, umbono wekhompyutha ushintsha indlela esisebenza futhi siphila ngayo. Amandla ayo okucubungula idatha yesikhathi sangempela ayenza ibe yinto ebalulekile ezinhlelweni zokusebenza ezifana nokutholwa kwezinto, izithombe zezokwelapha, kanye nezimoto ezizimele.

Kungani umbono womshini ubaluleke kangaka? Kuqhuba ukusebenza kahle kanye nokunemba lapho amehlo omuntu engaphumeleli khona. Isibonelo, ekukhiqizeni, izinhlelo zokuhlola okubonakalayo zithola amaphutha amancane, zisindisa izigidi ekukhunjweni. Kwezokunakekelwa kwempilo, izithombe zezokwelapha eziqhutshwa amanethiwekhi e-convolutional neural zithola umdlavuza kusenesikhathi, zithuthukise imiphumela yeziguli. Okwamanje, izimoto ezizimele zithembele ekulandeleni ukunyakaza kanye nokuqashelwa kwamaphethini ukuze zihambe ngokuphephile, zinciphise izingozi. Lokhu kuthuthuka kuvela emashumini eminyaka ocwaningo, ukuhlanganisa ukuhlukaniswa kwezithombe, ukuqashelwa kobuso, kanye nokubona kwe-3D ukubhekana nezinselele zomhlaba wangempela.

Lesi sihloko sigxila ekusetshenzisweni okuphezulu kombono womshini, sibonisa umthelela wawo kuyo yonke imikhakha. Sizohlola ukuthi i-Industry 4.0 isebenzisa kanjani amarobhothi kanye nokuhlolwa okuzenzakalelayo, ukuthi abathengisi bathuthukisa kanjani ukuhlaziywa kokuziphatha kwamakhasimende, nokuthi ukuphathwa kwethrafikhi kuthuthukisa kanjani ngokuqashelwa kwezimpawu zomgwaqo. Lindela ukuqonda ngezitayela ezisezingeni eliphezulu njenge-AI ekhiqizayo kanye namamodeli olimi lombono, asekelwe emithonjeni efana ne-Science.gov kanye ne-IoT Analytics.

Umgomo wethu? Ukuhlaziya ukuthi umbono womshini uguqula kanjani izimboni futhi uvuse ukusungula izinto ezintsha. Kungakhathaliseki ukuthi ungumholi webhizinisi, umcwaningi, noma umthandi wezobuchwepheshe, uzothola izimo zokusetshenziswa ezingokoqobo kanye namathuba esikhathi esizayo.

II. Umbono Womshini Kwezempilo

Umbono womshini uguqula ukunakekelwa kwezempilo ngokuthuthukisa ukunemba kokuxilonga nokwenza lula ukunakekelwa kweziguli. Ngokusebenzisa umbono wekhompyutha nokufunda okujulile, izinhlelo zihlaziya izithombe zezokwelapha ngokunemba okungakaze kubonwe, okubamba izinkinga amehlo abantu angase angaziboni. Kusukela ekutholeni umdlavuza osesigabeni sokuqala kuya ekuqapheni ukululama, ukuhlaziya okubonakalayo okunikezwa amandla amanethiwekhi ezinzwa eziguquguqukayo kusindisa izimpilo.

Ekuxilongweni, Ukucutshungulwa kwesithombe kuyakhanya. Ama-algorithms okuskena ama-CT scan, ama-MRI, kanye nama-X-ray ukuhlonza ukukhubazeka okufana nezimila noma amaphethini e-COVID-19 kuma-X-ray esifubeni. Isibonelo, izifundo ezivela ku-PMC zibonisa amamodeli e-AI afinyelela ukunemba okungaphezu kuka-90% ekutholakaleni komdlavuza webele, edlula izindlela zendabuko. Lesi sivinini nokuthembeka kusho ukungenelela okusheshayo.

Izinzuzo zokuhlinzwa nazo. Iqiniso elingeziwe lihlanganisa izithombe zezokwelapha zesikhathi sangempela ngesikhathi sezinqubo, liqondisa odokotela abahlinzayo ngokunemba okuqondile. Amathuluzi afana nokuhlukaniswa kwezithombe aqokomisa izindawo ezibalulekile, anciphisa izingozi ekusebenzeni okungangenisi kakhulu. Okwamanje, ukulandelela ukunyakaza kuqapha ukunyakaza kweziguli kude, kusiza ukulandelela ukululama ngaphandle kokuvakashela izibhedlela. Cabanga ngokuhlolwa kokuphulukiswa kwenxeba ngezinhlelo zokusebenza ze-smartphone—ezilula futhi ezingabizi kakhulu.

Umphumela? Izibhedlela zinciphisa ukubambezeleka kokuxilonga, futhi iziguli zithola ukunakekelwa komuntu siqu. Kodwa izinselele ezifana nobumfihlo bedatha zisekhona, zifuna izinhlelo eziphephile zokuvikela amarekhodi abucayi.

 

III. Umbono Womshini Ezinhlelweni Zokuthutha Ezihlakaniphile

Umbono womshini uguqula ezokuthutha, wenza imigwaqo iphephe futhi ihlakaniphe ngokusebenzisa umbono wekhompyutha kanye nokufunda okujulile. Ngokusebenzisa izinhlelo zokuthutha ezihlakaniphile, kwenza kube lula ukuphathwa kwethrafikhi ngesikhathi sangempela, ukutholwa kwezinto, kanye nokushayela ngokuzimela, kushintshe indlela esihamba ngayo.

Ukuze kuphephe ithrafikhi, ukucubungula izithombe kuyaphumelela. Ama-algorithm afana ne-YOLO kanye ne-Faster R-CNN athola izimoto, abahamba ngezinyawo, kanye nezimpawu zomgwaqo ngokunemba okuphezulu. Ukuqashelwa kwezimpawu zomgwaqo kuqinisekisa ukuthi abashayeli nezinhlelo ezizimele bahlala benolwazi, kuyilapho ukutholakala kwezibani zomgwaqo kuvimbela ukushayisana. Izifundo ezivela ku-PMC ziqokomisa izinhlelo ezifinyelela ukunemba okungu-95% ezindaweni zasemadolobheni, zinciphisa amazinga ezingozi.

Izimoto ezizimele zithembele kakhulu ekuboneni komshini. Kusukela kusizo lomshayeli wezinga 1 kuya kuzinga 5 ukuzimela okugcwele, ukulandelela ukunyakaza kanye nokuqashelwa kwamaphethini kuhamba ezindaweni eziyinkimbinkimbi. Amakhamera ahlanganiswe namanethiwekhi e-convolutional neural akhomba izithiyo ngesikhathi sangempela, okwenza kube lula ukushintsha imizila ephephile noma ukuma okuphuthumayo. Izinkampani ezifana neTesla zisebenzisa umbono we-3D ukuze zigibele kahle.

Ukuqapha ithrafikhi nakho kuyazuzisa. Ukulandelela abahamba ngezinyawo kukhombisa ukuhamba ngezinyawo, kuyilapho ukubona okungahambi kahle kukhombisa izingozi noma udoti. Izinhlelo zokukhokha ezihlakaniphile zisebenzisa ukuhlukaniswa kwezithombe ukufunda amapuleti, zenza kube lula ukukhokha. Lawa mathuluzi athuthukisa ukuhamba, anciphise ukuminyana emadolobheni.

Izinselele ziyaqhubeka, njengokuqinisekisa ubumfihlo bedatha ekuqapheni nasekuphatheni isimo sezulu esibi esithinta ukuhlaziywa okubonakalayo. Noma kunjalo, umthelela awunakuphikwa—imigwaqo ephephile, ukuvalwa okuncane kwe-grid.

IV. Umbono Womshini Ekukhiqizeni Nasezimbonini 4.0

Umbono womshini ushintsha kakhulu ku-Industry 4.0, uqhuba ukusebenza kahle kanye nokunemba ekukhiqizeni. Ngokusebenzisa umbono wekhompyutha kanye nokufunda okujulile, amafektri afinyelela ukulawulwa kwekhwalithi okuhlakaniphile, ukwenza ngcono izinqubo, kanye nokuphathwa kwempahla, ukunciphisa izindleko kanye nokukhulisa umkhiqizo.

Ukuhlolwa okubonakalayo kuholela ekuhlaseleni. Amanethiwekhi e-neural aguquguqukayo askena imikhiqizo ukuze athole amaphutha afana nokuklwebheka noma ukungalungi kahle, okudlula kakhulu ukunemba komuntu. Ngokungafani nezinhlelo ezisekelwe emithethweni, ukucubungula izithombe okuqhutshwa yi-AI kuyavumelana namaphutha ahlukahlukene, kubambe izinkinga kusenesikhathi. I-IoT Analytics ibika ukuthi ukulawulwa kwekhwalithi okuzenzakalelayo kusindisa izitshalo kufika ku-$172M minyaka yonke ngokunciphisa ukubuyiselwa emuva.

Ukuthuthukiswa kwenqubo kungenye into ephumelelayo. Ukuhlaziywa kwesikhathi sangempela kuqapha imishini, ukubona ukuguguleka noma ukungasebenzi kahle ngokuhlukaniswa kwezithombe. Ukulungiswa okubikezelayo, njengokutholwa kokugqwala, kuvimbela ukuwohloka, kwandise impilo yomshini. Le ndlela eqhutshwa idatha igcina ukukhiqizwa kuzwakala kahle, kunciphisa isikhathi sokungasebenzi.

Ezindaweni zokugcina impahla, umbono womshini wenza kube lula ukuthutha. Ukulandelela impahla ngokutholwa kwezinto kuqinisekisa ukubalwa kwesitoko okunembile, kuyilapho i-OCR inquma amabhakhodi ukuze kuhlungwe kahle. Ama-drone ahlonyiswe ngombono we-3D, asetshenziswa yiziqhwaga ezifana ne-Amazon, askena amashalofu ngesikhathi sangempela, anciphisa umsebenzi wezandla. Lawa mathuluzi athuthukisa ukusebenza kahle kweketanga lokuhlinzeka, ahlangabezane nesidingo ngokushesha.

Izinselele ezifana nezindleko zokusetha eziphakeme kanye nobumfihlo bedatha zidinga ukuxazululwa, kodwa izinzuzo zicacile: ukusebenza okulula kanye nemikhawulo yokuncintisana.

V.Isigaba Sombono Womshini Kwezicelo Zokuthengisa Nezomthengi

Umbono womshini ushintsha ukuthengisa, uhlanganisa umbono wekhompyutha kanye nokufunda okujulile ukuze kuthuthukiswe okuhlangenwe nakho kwamakhasimende futhi kuthuthukise ukusebenza. Kusukela ekuthengeni okwenziwe ngezifiso kuya ezitolo ezihlakaniphile, ukuhlaziya ukuthengisa okuqhutshwa ukucubungula izithombe kushintsha imboni.

Kumakhasimende, ukubonwa kobuso kuletha okuhlangenwe nakho okuklanyelwe wona. Amakhamera asesitolo ahlaziya ukuziphatha kwamakhasimende, avumela ukukhushulwa okuqondiwe. Izinhlelo zokuhlola ezingokoqobo, zisebenzisa ukuhlukaniswa kwezithombe, zivumela abathengi ukuthi bahlole izimonyo noma izibuko ngedijithali, okuthuthukisa ukuthengisa. Amathuluzi okuhlaziya isikhumba, njengalawo avela ku-L'Oréal, asebenzisa ukuhlaziywa okubonakalayo ukuncoma imikhiqizo, okuthuthukisa ukwaneliseka.

Imisebenzi yesitolo ithola ukusebenza kahle ngokubona komshini. Amamephu okushisa avela ekulandeleleni ukunyakaza aveza izindawo ezinabantu abaningi, okwenza kube ngcono ukwakheka. Ukutholwa kokwebiwa ngezinhlelo zokuqapha kunciphisa ukulahlekelwa, kanti i-AI ibona ukunyakaza okusolisayo ngesikhathi sangempela. Umbiko we-IoT Analytics ka-2023 uphawula ukuthi abathengisi bonga u-15% ekuncipheni besebenzisa ukutholwa kwezinto.

Ukuphathwa kwempahla nakho kuyakhanya. Ukuqashelwa kwezithombe kulandelela amazinga esitoko, kunciphisa ukugcwala kwesitoko. I-OCR iskena amalebula, isheshise ukuqoqwa kabusha kwezimpahla. Lawa mathuluzi aqinisa uchungechunge lokuhlinzeka, aqinisekise ukuthi imikhiqizo iyahlangabezana nesidingo. Izinhlelo ze-Walmart ezisekelwe embonweni, isibonelo, zivumelanisa amashalofu nama-oda aku-inthanethi kalula.

Izinselele ezifana nobumfihlo bedatha zidinga izindlela zokuphepha eziqinile, kodwa inzuzo isobala: amakhasimende ajabule kakhulu, imisebenzi esezingeni eliphansi.

VI. Izitayela Ezivelayo Neziqondiso Zesikhathi Esizayo

Umbono womshini ushintsha ngokushesha, ukhuthazwa yintuthuko ekuhlakanipheni kokwenziwa kanye nokuthuthuka kwehadiwe. Lezi zindlela zithembisa ukwandisa izinhlelo zokusebenza, kusukela kumarobhothi ahlakaniphile kuya embonweni wekhompyutha ofanele, okwakha ikusasa eliguquguqukayo.

Ukusungula izinto ezintsha zokufunda okujulile kuhola indlela. Ukwakheka kwe-transformer, njengalezo ezikumamodeli olimi lokubona, kuhlanganisa umbhalo nezithombe zemisebenzi efana nokuphendula imibuzo ebonakalayo. I-AI ekhiqizayo, kufaka phakathi amamodeli okusabalalisa kanye nama-GAN, idala idatha yokwenziwa, inciphisa ukushoda kwesethi yedatha. Ucwaningo lwe-Springer lwango-2023 luphawula ukuthi la mamodeli anciphisa izindleko zokuqeqesha ngo-20%, okwandisa ukusebenza kahle kokucubungula izithombe.

Ukuthuthukiswa kwehadiwe kukhulisa amakhono. Amakhamera okubona komshini amancane, afanelekela ukuthwebula izithombe zesayensi, aletha idatha enesisombululo esiphezulu ngezindlela ezihlanganisiwe. Ukusetshenziswa kwe-FPGA kusheshisa ukuhlaziya kwesikhathi sangempela, okubalulekile ezimotweni ezizimele kanye ne-Industry 4.0. Ama-chipset amasha athuthukisa amanethiwekhi e-neural convolutional, okwenza izinhlelo zisheshe futhi zishibhile.

Izinselele zokuziphatha zinkulu. Izinkinga zobumfihlo bedatha, ikakhulukazi ekubonweni kobuso, zifuna imithetho eqinile. Izinhlelo zokuqinisekisa i-biometric kumele zilinganise ukuphepha nokwethenjwa kwabasebenzisi. Okwamanje, ukwenza ngokuzenzakalelayo kuyingozi yokususa imisebenzi yamakhono aphansi, kudinga ukuqeqeshwa kabusha kwabasebenzi, njengoba kuqokomisa i-IoT Analytics.

Lezi zindlela zikhomba ikusasa lapho umbono womshini uhlangana khona kalula empilweni yansuku zonke, kusukela ekuthathweni kwezithombe zezokwelapha kuya ekuphathweni kwethrafikhi.

VII. Isiphetho

Umbono womshini uvele njengesikhungo esinamandla, uqhuba ushintsho embonini kuzo zonke ezempilo, ezokuthutha, ezokukhiqiza, kanye nezokuthengisa. Ngokusebenzisa umbono wekhompyutha, ukufunda okujulile, kanye nokucubungula izithombe, kunikeza izixazululo zombono womshini ezithuthukisa ukusebenza kahle kanye nokunemba. Kwezokunakekelwa kwempilo, ukuthathwa kwezithombe kwezokwelapha kuthuthukisa ukunemba kokuxilonga, kusindisa izimpilo. Izimoto ezizimele zithembele ekutholweni kwezinto ukuze zithole imigwaqo ephephile, kuyilapho i-Industry 4.0 isebenzisa ukuhlolwa okubonakalayo ukuze inciphise izindleko. Izitolo zisebenzisa ukuqashelwa kobuso kanye nokuhlaziywa kokuthengisa ukuze ziphakamise okuhlangenwe nakho kwamakhasimende, okufakazela ukuguquguquka kobuchwepheshe.

Lokhu akusikho nje ukuduma kobuchwepheshe—kuwumthelela wangempela. Amanethiwekhi e-Convolutional neural kanye namamodeli olimi lombono axazulula izinkinga eziyinkimbinkimbi, kusukela ekuphathweni kwethrafikhi kuya ekulandeleni isitokwe. Noma kunjalo, izinselele ezifana nobumfihlo bedatha kanye nokufuduka kwemisebenzi kudinga izindlela zokuziphatha ze-AI. Ukusungula izinto ezintsha okunomthwalo wemfanelo kuqinisekisa ukuthi umbono womshini uzuzisa wonke umuntu, kulinganisela inqubekela phambili nokwethenjwa.

Ikusasa liqhakazile kodwa lidinga isenzo. Amabhizinisi kufanele ahlole ukuhlaziya kwesikhathi sangempela kanye ne-AI ekhiqizayo ukuze ahlale encintisana. Abacwaningi bangadlula imingcele ngezakhiwo ze-transformer, kuyilapho abenzi bezinqubomgomo kumele babeke phambili ubumfihlo bedatha. Njengoba i-IoT Analytics iphawula, izinkampani ezisebenzisa izixazululo zombono womshini zibona izinzuzo zokusebenza kahle ezifika ku-25%, isizathu esibalulekile sokuthatha isinyathelo.

Ukuze kusetshenziswe, kubanjiswane nabanolwazi abakhiqizi bamakhompyutha afakiwe kubalulekile, ikakhulukazi lapho kusetshenziswa izixazululo ezifana I-AFE-R770 kwezicelo zezinga lezimboni. ikhompyutha efakwe ezimbonini ingabhekana nezimo ezidinga umzamo omkhulu, inikeza ukuthembeka nokusebenza okuqhubekayo. Amadivayisi afana ne- I-Mini PC Industrial J1900 zilungele ukucubungula umkhawulo, kuyilapho zisebenza kahle kakhulu I-PC yebhokisi elifakiwe inganika amandla imithwalo yemisebenzi ye-AI ethuthukisiwe ensimini.