Google yana fitar da bayanai da samfurin koyon injin don raba sautuna

Google aka buga ma'aunin bayanai na bayanai gauraye masu sauti waɗanda za a iya amfani da su a cikin tsarin koyo na na'ura da ake amfani da su don raba gauraye sautunan sabani a cikin sassansu ɗaya. An kuma buga samfurin koyo mai zurfi na inji (TDCN++) wanda za'a iya amfani dashi a cikin Tensorflow don raba sautuna. Bayanan da aka shirya bisa tarin freesound.org и buga lasisi ƙarƙashin CC BY 4.0.

Shirin FUSS (Free Universal Sound Separation) da aka gabatar yana da nufin magance matsalar raba kowane adadin sauti na sabani, wanda ba a san yanayinsa a gaba ba. Sauran tsarin makamantan su gabaɗaya sun iyakance ga aikin bambancewa tsakanin wasu sautuna, kamar muryoyi da waɗanda ba surutu ba, ko mutane daban-daban masu magana.

Database ya ƙunshi game da 20 dubu mixings. Kit ɗin ya haɗa da martanin da aka riga aka ƙididdige ɗaki ta amfani da na'urar na'urar na'urar da aka gina ta al'ada wacce ke yin la'akari da tunanin bango, wurin tushen sauti, da wurin makirufo.

source: budenet.ru