This project is a Turkish Sign Language Alphabet Recognizer using Python. You can make a sentences and make it talk.
 |
 |
 |
 |
 |
|---|---|---|---|---|
| A | B | C | D | E |
 |
 |
 |
 |
 |
|---|---|---|---|---|
| F | G | H | I | K |
 |
 |
 |
 |
 |
|---|---|---|---|---|
| L | M | N | O | P |
 |
 |
 |
 |
 |
|---|---|---|---|---|
| R | S | T | U | V |
 |
 |
 |
 |
 |
|---|---|---|---|---|
| Y | Z | Space | Backspace | Dot |
pip install -r requirements.txt
You can download trained data from Here
I tried diffirent models and i implemented the most optimum model in the model.py file. You can compare models with tensorboard.
tensorboard --logdir=logs
First of all you need to train the model and save it.
python model.py
Then you can use this model to recognize sign language in real-time through webcam.
python prediction.py
Also you can create your own dataset with collectImg.py file. Once you done that don't forget convert images to numpy array with createData.py. Be careful with the file paths.
If you want to use Turkish voice, you have to follow this steps on Windows. (Alternative Turkish resource)