Ii's a prototype that translate 10 signs: ["me","close","continue","eat","Where","listen","open","short","tall","watch"]
In arabic: ["أنا","يغلق","يستمر","يأكل","أين","يسمع","يفتح","قصير","طويل","يشاهد"]
You can Upload the image and press "Go" button. then you will git the sign class into text and voice.
I collected the data myself, there are about 18805 images.
I used transfer learning with ResNet-50 You can download the 'weights.h5' file from here => https://drive.google.com/file/d/1-6a6rsUF-EqwOumYNEgBA8vLpjWhPmUB
I built a single page using HTML, CSS, JS and the backend with Flask.