
American Sign Language Interpreter Using an RGB-D Camera and Machine Learning
Software as a Service
NoCategory
AI & Machine LearningClients
Mapua University
Techstack
PythonTensorFlowMediaPipeOpenCV3DCNNLSTM
Purpose
Identify and classify American Sign Language hand gestures from RGB-D camera input, using 3DCNN and LSTM deep learning models to improve communication and accessibility for ASL users.
Description
The pipeline extracts hand landmarks from RGB-D video using MediaPipe and OpenCV, then feeds the sequences to an ensemble of 3D Convolutional Neural Network and Long Short-Term Memory models built in TensorFlow. Per-gesture models are trained and evaluated with confusion-matrix analysis, and a threaded webcam pipeline runs gesture recognition in real time. Developed as a thesis project at Mapua University.




