Cat Breed Auto Identify Photo

Siyini isicelo?
Icacisa uhlobo lwekati ngezithombe usebenzisa ikhamera yedivayisi yakho noma igalari yesithombe.

Isebenza kanjani?
Isithombe sinikezwa okokufaka kwenethiwekhi ye-neural (okwamanje kusetshenziswa i-architecture ye-EfficientNetV2) futhi ekuphumeni kwaso kwakhiwa i-hypothesis mayelana nokuthi yiluphi uhlobo lwekati oluboniswa kulesi sithombe. Inguqulo entsha ye-classifier ayisadlali futhi isabela kuphela ezithombeni zamakati angempela. Amakati adwetshiwe, opopayi, amathoyizi, izinja, ezinye izilwane, izithombe zabantu - inethiwekhi ye-neural evame ukungazinaki.

Kuyini ukunemba kokuqashelwa?
Lolu hlelo luqeqeshelwe ukubona izinhlobo zamakati ezingama-62 ezithombeni eziyi-13,000. Kule nguqulo yohlelo lokusebenza, ukunemba kokuqashelwa kohlobo lwamakati kwaba ngu-63% ezithombeni eziyizinkulungwane ezimbili ezivela kusampula yokuhlola (ezingasetshenzisiwe ekuqeqesheni isigaba) kanye nama-86% kuzo zonke izithombe ezitholakalayo. Isizindalwazi sokuqeqeshwa sezithombe zamakati siyengezwa futhi siyathuthukiswa, ngakho inani lezinhlobo kanye nekhwalithi yokuqashelwa kwazo kuzokhuphuka ekukhishweni okusha.

Imigomo Yekusasa.
Izongezwa ukwengeza isethi yokuqeqeshwa yezithombe zekati izibonelo zakho futhi ngaleyo ndlela ikhulise ngokuqhubekayo inani lezinhlobo zamakati nokunemba kokuqashelwa. Inhloso yephrojekthi ukudala uhlelo lochwepheshe olukwazi ukubona izithombe zonke izinhlobo zamakati ezaziwayo.