How rrupeshh/simple-sign-language-detector works
Project type and purpose
• This is a standalone machine-learning pipeline (not a web app) for American Sign Language (ASL) recognition.
• It uses Python scripts and a Convolutional Neural Network (CNN) in Keras/TensorFlow, plus OpenCV for image capture and preprocessing.
File structure
• Top-level scripts:
• .py – captures hand images via webcam, writes to mydata/training_set or mydata/test_set
• cnn_model.py – defines and trains the CNN on images in mydata/training_set; saves Trained_model.h5
• test_model.py – loads the trained model, evaluates it against mydata/test_set
• recognise.py – real-time inference: reads webcam frames, segments the hand (via HSV trackbars), predicts ASL letters
• Data directories:
• mydata/training_set/{A…Z} and mydata/test_set/{A…Z} – labeled image folders
• Model artifact:
• Trained_model.h5 – serialized Keras model
README for additional insights
• Confirms tools: Python3, OpenCV3, TensorFlow, Keras
• Workflow: capture → train → test → recognise
• No microservices or external APIs—pure local execution
Elements
• Clear labels for each module/script and folder