Wave at your laptop and it waves back, sort of. Sign Language Detection watches your hand through the webcam and translates the gesture into the letter you’re signing, live, frame by frame.
Under the hood it’s a full machine learning pipeline. First it collects training samples straight from the camera, then it trains a neural network on those gestures, and finally it predicts in real time with the guess drawn right back onto the video feed. A separate evaluation step keeps the model honest by measuring how often it actually gets the letter right.
The interesting part is that none of this is pre-baked. The dataset is built by hand, gesture by gesture, so the model learns from your camera, your lighting, and your hands rather than someone else’s.
Built with Python, TensorFlow, Keras, and OpenCV.