Hand Gesture Recognition System

This project implements a real-time hand gesture recognition system using TensorFlow, OpenCV, and MediaPipe. The system captures video frames from the webcam, processes them to detect hands, and uses a pre-trained model to predict hand gestures. The predicted gestures are then displayed on the video feed.

Table of Contents

1. Requirements
2. Setup
3. Running the Application
4. Model
5. How It Works
6. Troubleshooting
7. Credits
8. Model Requests

Requirements

Before running the code, ensure you have the following packages installed:

• TensorFlow
• OpenCV
• MediaPipe
• Keras
• NumPy

You can install these packages using pip:

pip install tensorflow opencv-python mediapipe keras numpy

Setup

1. Clone the Repository:

   Clone the repository to your local machine.

   git clone https://github.com/yourusername/hand-gesture-recognition.git
   cd hand-gesture-recognition

2. Model File:

   Ensure you have the pre-trained model file jayanthModel.h5 in the appropriate directory. Update the path to the model file in the code if necessary.

   model = load_model(r"\Users\prash\OneDrive\Desktop\jay_projecf\jay_projecf\jayanthModel.h5")

3. Run the Application:

   Execute the Python script to start the hand gesture recognition system.

   python hand_gesture_recognition.py

Running the Application

1. Start Video Capture:

   The application captures video from the default webcam. Ensure your webcam is connected and working properly.

2. Hand Gesture Prediction:

   The application processes each frame to detect hands and predicts the corresponding gesture using the pre-trained model.

3. Display Predictions:

   The predicted gesture is displayed on the video feed in real-time.

4. Exit:

   Press the q key to exit the application.

Model

The model used in this project is a pre-trained Convolutional Neural Network (CNN) saved as jayanthModel.h5. The model is trained to recognize various hand gestures corresponding to letters in the alphabet.

The labels for the predictions are:

letterpred = ['A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'K', 'L', 'M', 'N', 'O', 'P', 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X', 'Y']

How It Works

1. Video Frame Capture:

   Captures frames from the webcam using OpenCV.

2. Hand Detection:

   Uses MediaPipe's Hands module to detect hand landmarks in the frame.

3. Region of Interest Extraction:

   Extracts the bounding box around the detected hand and preprocesses the region for prediction.

4. Gesture Prediction:

   Resizes the hand region to 28x28 pixels, converts it to grayscale, normalizes pixel values, and feeds it into the CNN model to get the predicted gesture.

5. Display Prediction:

   Displays the predicted gesture on the video frame using OpenCV.

Troubleshooting

• No Frame Capture: Ensure your webcam is connected and selected correctly. Check the cap = cv2.VideoCapture(0) line to ensure the correct device index is used.

• Invalid Model Path: Ensure the path to the jayanthModel.h5 file is correct.

• Dependencies: Ensure all required packages are installed. Use pip install to install any missing packages.

Credits

This project is developed by Jayanth Rahul as part of a guitar learning system. The code leverages TensorFlow, OpenCV, and MediaPipe to achieve real-time hand gesture recognition.

Model Requests

If you need the jayanthModel.h5 file for this project, please reach out, and I will provide it to you. Feel free to contact me for any questions or issues regarding this project!