PFEBOUIDI

BOUIDI Abderrahmane

Mayelana nalolu hlelo lokusebenza

Ukuqashelwa Kwesithombe, kumongo we-Computer Vision, yikhono lesofthiwe lokuhlonza izinto, izindawo, abantu, imibhalo kanye nezenzo ezithombeni. Amakhompyutha angasebenzisa ubuchwepheshe bokubona komshini ngokuhlanganiswa nekhamera kanye nesofthiwe yobuhlakani bokwenziwa ukuze azuze ukuqashelwa kwesithombe.
Ukuhlukaniswa kwesithombe kusho inqubo ekubonweni kwekhompyutha engahlukanisa isithombe ngokusekelwe kokuqukethwe kwaso okubukwayo. Isibonelo, i-algorithm yokuhlukanisa isithombe ingaklanywa ukubonisa ukuthi isithombe siqukethe noma cha umuntu. Nakuba ukutholwa kwezinto kuyinto encane kubantu, ukuhlukaniswa kwesithombe okuqinile kuseyinselele ezinhlelweni zokubona kwekhompyutha.
Inhloso yalolu cwaningo ukunquma ukuthi yini eyenza inethiwekhi ye-neural ejulile icubungule idatha eyinkimbinkimbi, njengedatha yesithombe/ividiyo, isheshe futhi inembe kakhudlwana, sizohlola izakhiwo zakamuva eziyimpumelelo zenethiwekhi ye-neural ukuze sinqume ukuthi iyiphi esebenza kahle kakhulu (futhi eshesha kakhulu) izakhiwo ekuhlukaniseni izithombe, futhi sizophinda sicwaninge ukuthi yimaphi amasu okuthuthukisa asebenza kangcono kulolu hlobo lwedatha.
Sizama ukuqonda ukuthi abacwaningi basanda kuthatha kanjani isinyathelo esikhulu phambili ekuqashelweni okubonakalayo ngokuhlukanisa izithombe, futhi sibone ukuthi bathole kanjani amaphuzu okunemba okumangalisayo kunselelo ye-ImageNet. Uma kucatshangelwa ukuthi singayicubungula kanjani idatha eyinkimbinkimbi njengedatha yesithombe ngokushesha, singayisingatha kanjani inkinga yokufakwa ngokweqile kule datha, futhi singasinciphisa kanjani isikhathi sokuqeqeshwa sezakhiwo zethu.