How najnin26/hand-gesture-recognition works
Project’s Core Type and Purpose
• This is a computer vision and machine learning project focused on real-time hand gesture (and sign language) recognition.
• It uses live video capture to detect, process, and classify hand gestures using OpenCV for image processing and a TensorFlow/Keras-based machine learning model (stored under the mp_hand_gesture folder).
File Structure
• Data Files & Assets:
• DIP-Hand Gesture_compressed.pdf (likely a design or background document)
• data.pickle (may store preprocessed features or training data)
• gesture.names (list of gesture labels available in the system)
• Source Code:
• datacollection.py – responsible for gathering or processing raw data (e.g., capturing hand gesture images).
• test.py – could be used for running tests or demonstrations.
• Model Files:
• The mp_hand_gesture directory includes a TensorFlow/Keras saved model, including keras_metadata.pb, saved_model.pb, and variables directory for model weights.
• Other Files:
• README.md provides an overview of the system’s functionality, input requirements, and the gesture names.
• The presence of a “new” directory (or file) suggests there may be an iterative or experimental component.
System’s High-Level Architecture Components
a. Input/Acquisition Component:
• Video Capture: Utilization of a webcam feed (or similar source) via OpenCV.
• Data Acquisition Script (datacollection.py): Responsible for capturing and possibly preprocessing raw hand gesture frames.
b. Preprocessing & Feature Extraction:
• Hand Detection: Process frames to isolate the hand region (could incorporate MediaPipe or OpenCV functionality).
• Feature Extraction: Identify key features (e.g., contours, finger positions) from the detected hand.
c. Gesture Classification Module:
• Machine Learning Model: The TensorFlow/Keras model stored in mp_hand_gesture is responsible for classifying the processed frames into one of the predefined gestures.
• Gesture Labels: Reference the gesture.names file that contains the list of gestures (e.g., "okay", "peace", etc.).
d. Output Component:
• Gesture Output: Final recognized gesture can be displayed on-screen and used to trigger corresponding actions or feedback.
e. (Optional) Data & Training Storage:
• Data Files: data.pickle may be used to store collected training data or intermediate processing results.
• Documentation/Reference Materials: DIP-Hand Gesture_compressed.pdf for design insights.
Relationships and Data Flow
• The video capture provides real-time frames to the input component.
• Frames are passed to the preprocessing module, where hand detection and feature extraction occur.
• The processed data is then forwarded to the gesture classification module (the TensorFlow model in mp_hand_gesture).
• Recognized gestures (output) are then displayed or sent to any interfacing components (possibly integrated in test.py or other parts of the application).