Ukuhlenga kabusha kwensimu yokuhlola isimo se-soya phakathi kwehlobo - okuqeda amandla, kepha umsebenzi odingekile lapho ususe izinhlobo ezintsha.
Abafuyi kufanele bazulazule nsuku zonke ngaphansi kwelanga elivuthayo ezikhathini ezibucayi zesikhathi sokukhula ukuze bathole izitshalo ezibonisa izici ezifiselekayo ezifana nokuvuthwa kokuqala kwama-pods. Kepha, ngaphandle kokuba nethuba lokusebenzisa ukutholwa kwalezi zibonakaliso, ososayensi abakwazi ukuvivinya amasayithi amaningi njengoba bengathanda ukukhulisa isikhathi sokuqeda izinhlobo ezintsha emakethe.
Ekutadisheni okusha kwe-University of Illinois, ososayensi babikezela isikhathi sokuvuthwa kwe-soya kungakapheli izinsuku ezimbili besebenzisa izithombe ezivela ku-Drones kanye ne-intelligence.
"Ukuhlolwa kokuvuthwa kokuvuthwa kwe-POD kudinga isikhathi esiningi futhi lapha kuvame ukwenze iphutha, ngoba lolu hlelo lokuhlola lususelwa kumbala we-pod, futhi kukhona ubungozi bokukunquma ngokungafanele," kusho uNicholas Martin , Uprofesa ohambisana noMnyango Wezokuvumela e-Illinois kanye noMthethosisekelo wesifundo. "Abaningi bazame ukusebenzisa ama-Snapshots basuka e-Drones ukuyohlola ukuvuthwa, kepha singowokuqala ukuthola indlela efanele yokukwenza."
URodrigo Trevizan, umfundi osebenza ngodokotela osebenza noMartin, wafundisa amakhompyutha ukuthola izinguquko zombala ezithombeni ezivela ku-Drones eziqoqwe ezilingweni ezinhlanu, isizini ezintathu ezikhulayo namazwe amabili. Kubalulekile ukuqaphela ukuthi amakhompyutha akwazi ukucubungula futhi ahumushe ngisho nezithombe "ezimbi".
"Ake sithi sifuna ukuqoqa izithombe njalo ezinsukwini ezintathu, kodwa lapho amafu avele noma anane, okuthinta ikhwalithi yezithombe. Ekugcineni, lapho uthola idatha yeminyaka ehlukene noma ezindaweni ezahlukene, bonke bazobukeka behlukile ngokubuka kwenani lezithombe, ngezikhathi ezithile nokunye. Ukusungulwa okuyinhloko esikukhethile yindlela esingabheka ngayo yonke imininingwane etholakele. Imodeli yethu isebenza kahle noma ngabe idatha yayihamba kaningi kangakanani, "kusho uTrevizan.
I-Trevisan yasebenzisa uhlobo lobuhlakani bokufakelwa, okubizwa ngokuthi amanethiwekhi ajulile we-congdenueling neural neural (CNN). Uthi i-CNN ifana nendlela umuntu wobuchopho bomuntu afunda ukuhumusha izingxenye zezithombe - umbala, ukwakheka, ukuthungwa - okuwukuthi, imininingwane etholakala emehlweni ethu.
"I-CNN ithola izinguquko ezincane ngombala, ngaphandle kwamafomu, imingcele kanye nokwenziwa. Kithina, okubaluleke kakhulu kwawumbala. Kepha inzuzo yamamodeli wobuhlakani bokufakelwa, esasisebenzisa, ukuthi kuzoba lula impela ukusebenzisa imodeli efanayo ukubikezela esinye isimilo, njengesivuno noma i-span. Ngakho-ke, manje njengoba sinalezi zimodeli, abantu kufanele babe lula kakhulu ukusebenzisa isu elifanayo ukufeza eminye imisebenzi eminingi, "kuchaza uTrevizan.
Ososayensi bathi ubuchwepheshe buzoba wusizo ikakhulukazi ekufuneni izinkampani zezohwebo.
"Sibe nabalingani bezinhlaka abahlanganyele esifundweni esizofuna ukuyisebenzisa eminyakeni ezayo. Futhi benza umnikelo omuhle kakhulu, obalulekile. Bebefuna ukuqiniseka ukuthi izimpendulo zisebenza abafuyi bensimu ezenza izinqumo ezikhetha izitshalo nabalimi, "kusho uNicholas Martin.
(Umthombo: Farmtario.com. Photo: Getty Izithombe).