A real-time sign language alphabet recognition system built with Python, MediaPipe, and a Random Forest classifier. The project is structured into three main phases: dataset creation and preprocessing, model training, and real-time inference using webcam input.
- Detects and classifies hand gestures representing alphabets.
- Real-time gesture recognition using webcam feed.
- Uses hand landmark detection for robust feature extraction.
- Custom dataset creation and preprocessing.
- Trained using the Random Forest classification algorithm.
- Visual feedback with prediction results displayed live.
- Image Collection: Hand gesture images are collected and organized into labeled directories by gesture.
- Feature Extraction: Uses MediaPipe to extract hand landmarks from images.
- Normalization: Features are normalized for consistency.
- Labeling: Each processed feature set is paired with the correct label.
- Storage: Processed features and labels are stored for training.
- Data Splitting: Dataset is split into training and testing sets.
- Training: A Random Forest classifier is trained on the extracted features.
- Evaluation: Performance is measured using accuracy, precision, recall, and F1-score.
- Model Saving: The trained model is saved for use in real-time inference.
- Live Input: Captures video from webcam.
- Feature Extraction: Processes live images to extract hand landmarks.
- Prediction: Classifies the gesture using the trained model.
- Feedback: Displays the predicted alphabet on the screen.
- Programming Language: Python
- Libraries: OpenCV, MediaPipe, Scikit-learn, NumPy, joblib
- Model: Random Forest Classifier
The system demonstrates:
- ✅ High accuracy in classifying alphabetic hand gestures
- ⚡ Fast real-time inference using webcam input
- 🔍 Reliable feature extraction with MediaPipe for landmark detection
- 📈 Consistent performance across various lighting conditions and hand positions
- 👁️ Visual feedback with recognized gesture displayed live on-screen