SignLanguageRecognition

An overview of the model

Setting up

Run calibration.py to calibrate your hand gestures for every alphabet. This will aid in improving the accuracies while predicting from the trained model.

It will save a numpy array where each element specifies whether the fingers are open/closed/half open-closed.

Detecting from web cam:

Reference:

This project uses a CNN model that follows AlexNet architecture.

Image credits to Krizhevsky et al., the original authors of the AlexNet paper.

Notes

Please note that detector_mediapipe.py gives more accurate results than detect_webcam.py.

The requirements.txt file contains all the dependencies needed for the project.

pip install -r requirements.txt

Mediapipe docs Trained models (Google Drive link)