This is computer vision; real-time hand gesture recognition software implemented in Python and TensoFlow. Computer vision is a subfield of artificial intelligence (AI) that enables computers to “see” by deriving meaningful information from digital images, videos and/or act based on their perception.
First, I used MediaPipe's pose and hands ML model to detect and extract fetures of my hands and body pose. On top of that model I built LSTM neural network that learns those fetures (landmarks/keypoints) and later recognises them. You can train then deep LSTM neural network with your own hand gestures, or reuse my pre-trained model stored in .H5 file.
Hand gestures that the .H5 pre-trained model can detect:
- Like 👍
- Ok 👌
- Hello 👋
Author: Armin Smajlagić