Andrey Savchenko daga reshen Nizhny Novgorod na Babban Makarantar Ilimin Tattalin Arziki ya wallafa sakamakon binciken da ya yi a fannin koyon na'ura da ke da alaka da fahimtar motsin zuciyar mutane a fuskokin mutanen da ke cikin hotuna da bidiyo. An rubuta lambar a Python ta amfani da PyTorch kuma tana da lasisi ƙarƙashin lasisin Apache 2.0. Akwai samfuran shirye-shiryen da yawa, gami da waɗanda suka dace don amfani akan na'urorin hannu.
Dangane da ɗakin karatu, wani mai haɓaka ya ƙirƙiri shirin sevimon, wanda ke ba ku damar bin canje-canje a cikin motsin rai ta amfani da kyamarar bidiyo da kuma taimakawa wajen sarrafa tashin hankali na tsokar fuska, alal misali, don kawar da wuce gona da iri, a kaikaice yana shafar yanayi kuma, tare da amfani na dogon lokaci. hana bayyanar wrinkles na fuska. Ana amfani da ɗakin karatu na CenterFace don tantance matsayin fuska a cikin bidiyo. An rubuta lambar sevimon a Python kuma tana da lasisi ƙarƙashin AGPLv3. Lokacin da kuka ƙaddamar da shi a karon farko, ana ɗora samfuran, bayan haka shirin baya buƙatar haɗin Intanet kuma yana aiki gaba ɗaya kai tsaye. An shirya umarnin ƙaddamarwa akan Linux/UNIX da Windows, da kuma hoton docker don Linux.
Sevimon yana aiki kamar haka: na farko, an gano fuska a cikin hoton kyamara, sannan an kwatanta fuska da kowane motsin motsi takwas (fushi, raini, kyama, tsoro, farin ciki, rashin jin dadi, bakin ciki, mamaki), bayan haka wani takamaiman. An ba da maki kamance don kowane motsin rai. Ana adana ƙimar da aka samu a cikin log in tsarin rubutu don bincike na gaba ta shirin sevistat. Ga kowane motsin rai a cikin fayil ɗin saituna, zaku iya saita iyakoki na sama da ƙasa, idan an ketare, ana ba da tunatarwa nan da nan.
source: budenet.ru