Unredacter, kayan aiki don gano rubutun pixel, an gabatar da shi

An gabatar da kayan aikin Unredacter, wanda ke ba ku damar dawo da ainihin rubutun bayan ɓoye ta ta amfani da matattara dangane da pixelation. Misali, ana iya amfani da shirin don gano mahimman bayanai da kalmomin shiga masu ƙima a cikin hotunan kariyar kwamfuta ko hotunan hotuna. An yi iƙirarin cewa algorithm ɗin da aka aiwatar a cikin Unredacter ya fi sama da makamantan abubuwan amfani da aka samu a baya, kamar Depix, kuma an yi nasarar amfani da shi don cin jarrabawar gano rubutun pixilated wanda dakin gwaje-gwaje na Jumpsec ya gabatar. An rubuta lambar shirin a cikin TypeScript kuma an rarraba ta ƙarƙashin lasisin GPLv3.

Don mayar da rubutu, Unredacter yana amfani da hanyar zaɓi na baya, gwargwadon abin da aka kwatanta sashe na ainihin hoton pixelated tare da bambance-bambancen da aka haɗa ta hanyar bincike ta nau'i-nau'i na haruffa masu pixel tare da canje-canje daban-daban da canza halaye. Yayin binciken, zaɓin da ya fi dacewa ya yi daidai da guntun asali ana zaɓe a hankali. Don yin aiki cikin nasara, kuna buƙatar yin la'akari daidai girman, nau'in da sigogin shigar da font ɗin, kazalika da lissafin girman tantanin halitta a cikin grid pixelation da matsayi na grid mai rufi akan rubutu (zaɓuɓɓukan grid diyya ana tsara su ta atomatik) .

Unredacter, kayan aiki don gano rubutun pixel, an gabatar da shi

Bugu da ƙari, za mu iya lura da aikin DepixHMM, a cikin tsarin da aka shirya wani nau'i na kayan aiki na Depix, wanda aka fassara zuwa wani algorithm dangane da samfurin Markov mai ɓoye, godiya ga wanda zai yiwu don ƙara daidaito na sake gina alamar.

source: budenet.ru

Add a comment