How muzammil-ansari-17/signlanguagetranslator works
This is a webcam-based ASL gesture translator. The main workflow captures frames, detects and classifies hand landmarks, builds words and sentences, offers word suggestions, and can translate text or speak it. A separate model-preparation workflow extracts landmarks from image classes and trains the classifier used at runtime. The README describes real-time detection and speech; sampled code confirms these paths, though the UI currently enables only English. Training and translation-service wiring are based on the sampled scripts and modules.
Main entry points: main.py App loop and UI · gesture_detector.py Gesture detector · train_model.py Model training · tts_engine.py Speech synthesis
Components 18
Translation app
- App loop and UI
main.py - Word suggestions
word_logic.py
Gesture recognition
- Gesture detector
gesture_detector.py - Hand landmarks
gesture_detector.py - Gesture classifier
data/asl_model.h5 - Label mapping
data/label_encoder.pkl
Model preparation
- Training images
data/asl_alphabet_train - Landmark extraction
extract_landmarks.py - Landmark dataset
data/landmarks.csv - Model training
train_model.py
Language output
- Text translation
translator.py - Speech synthesis
tts_engine.py
Other components
- User
- Webcam
- Display
- Audio output
- Translation provider
- gTTS service
Connections 18
- User → App loop and UI: operates
- Webcam → App loop and UI: sends frames
- App loop and UI → Gesture detector: classifies frames
- Gesture detector → Hand landmarks: extracts landmarks
- Gesture detector → Gesture classifier: predicts gestures
- Gesture detector → Label mapping: decodes labels
- App loop and UI → Word suggestions: requests suggestions
- App loop and UI → Text translation: translates text
- Text translation → Translation provider: requests translation
- App loop and UI → Speech synthesis: requests speech
- Speech synthesis → gTTS service: generates audio
- Speech synthesis → Audio output: plays audio
- App loop and UI → Display: renders interface
- Training images → Landmark extraction: provides images
- Landmark extraction → Landmark dataset: writes landmarks
- Model training → Landmark dataset: reads landmarks
- Model training → Gesture classifier: saves model
- Model training → Label mapping: saves labels