Vision2Speech
Author: Szymon Dyrszka, Kevin Kozłowski, Oliwier Wilim, Marcik Lasak, Artur Hausman, Krzysztof Smyczek, Dominik Sadłocha
Institution: King John III Sobieski School Complex no. 6, Vocational Education Center in Jastrzębie-Zdrój
Country: Poland
e-mail: szdyrszka@gmail.com; kevkoz2207@gmail.com; oliwier.wilim@onet.pl
Vision2Speech translates sign language into speech in real time. Worn as smart glasses on a Raspberry Pi 5, it uses MediaPipe hand-tracking and a custom GRU neural network to recognize gestures from camera input, then speaks the corresponding phrase through a Bluetooth speaker — fully offline, giving deaf sign-language users an instant voice in everyday conversations.
