Machine Learning · 2022
Sign Language to Text Converter
Real-time action-detection translating sign language into text, using an LSTM over pose keypoint sequences.
My final-year machine learning project. Static image classification isn’t enough for sign language because the meaning is in the movement, so the model runs an LSTM over sequences of extracted pose and hand keypoints rather than classifying individual frames.
An iOS port followed, which is what got me interested in on-device inference and led to Drawing Guesser.