Open Internet by MindsNet

Improving ASL Text-to-Speech with Robust Hand Sign Recognition

The author is struggling with a clunky hand sign recognition system that requires holding a sign for 1.5 seconds to work. The current approach relies on frame-by-frame hand landmarks, which is unreliable. The author wants to migrate to a Vision LSTM approach but faces challenges with missing frames and model performance.

Computing & Technology, Computer Science, Artificial Intelligence