Wonke umuntu ofuye isilwane ekhaya uyasithanda. Nokho, ingabe inja yakho oyithandayo izobukeka inhle nakakhulu uma bekuwuhlobo oluhlukile? Ngenxa yethuluzi elisha elivela ku-NVIDIA elibizwa ngokuthi i-GANimals, ungakwazi ukuhlola ukuthi isilwane sakho osithandayo besizobukeka sisihle nakakhulu yini ukube bekuyisilwane esihlukile.
Ngasekuqaleni kwalo nyaka, i-NVIDIA Research Abasebenzisi be-inthanethi ngethuluzi lakhe le-GauGAN, elimvumele ukuthi aguqule imidwebo enzima ibe cishe yizithombe ze-photorealistic. Leli thuluzi belidinga abasebenzisi ukuthi bacacise ukuthi yiziphi izingxenye zesithombe okufanele zibe amanzi, izihlahla, izintaba nezinye izimpawu zendawo ngokukhetha umbala webhulashi ofanele, kodwa i-GANimals isebenza ngokuzenzakalelayo ngokuphelele. Okufanele ukwenze nje ukulayisha isithombe sesilwane sakho esifuywayo, futhi kuzodala uchungechunge lwezithombe ze-photorealistic zezinye izilwane ezigcina "isimo sobuso" sesampula.
Kuleli sonto, ephepheni elethulwe eNgqungqutheleni Yamazwe Ngamazwe Nge-Computer Vision eSeoul, eKorea, abacwaningi bachaze i-algorithm abayithuthukisile - . Imele Ukuhunyushwa kwesithombe-kuya-esithombeni okungagadiwe. Uma usebenzisa ubuhlakani bokwenziwa ukuguqula izici zesithombe esiwumthombo sibe isithombe esiqondiwe, ubuhlakani bokwenziwa ngokuvamile budinga ukuqeqeshwa ngeqoqo elikhulu lezithombe eziqondiwe ezinamaleveli okukhanya ahlukene nama-engeli ekhamera ukuze kukhiqizwe imiphumela ebukeka ingokoqobo. Kodwa ukwakha isizindalwazi esikhulu kangaka sesithombe kuthatha isikhathi eside futhi kukhawulela amandla enethiwekhi ye-neural. Uma i-AI iqeqeshelwe ukushintsha izinkukhu zibe ama-turkeys, yilokho kuphela okuzokwenza kahle.
Uma kuqhathaniswa, i-algorithm ye-FUNIT ingaqeqeshwa kusetshenziswa izithombe ezimbalwa zesilwane esiqondiswe kuso lapho ijwayezwa ngokuphindaphindiwe. Uma i-algorithm isiqeqeshwe ngokwanele, idinga kuphela isithombe esisodwa somthombo nezilwane eziqondiwe, ezingaba ngokungahleliwe ngokuphelele futhi ezingakaze zicutshungulwe noma zihlaziywe ngaphambilini.
Labo abanentshisekelo bangazama i-GANanimals ku , kodwa kuze kube manje imiphumela iyisinqumo esiphansi futhi ayifanele lutho ngaphandle kwezinjongo zemfundo noma ukwanelisa ilukuluku. Abacwaningi banethemba lokuthi ekugcineni bazothuthukisa amakhono e-AI kanye ne-algorithm ukuze maduze kwenzeke ukushintsha ubuso babantu ngaphandle kokuthembela kudathabheyisi enkulu yezithombe ezikhethwe ngokucophelela.
Source: 3dnews.ru