Real-Time Digit Detection with CNN

This project is a real-time digit detection application using a Convolutional Neural Network (CNN). The application captures images from a webcam, preprocesses them, and predicts the digit displayed in real-time.

Screenshots

Overview

This project demonstrates the application of deep learning techniques to detect handwritten digits in real-time using a webcam. The CNN model is trained on a custom dataset of digit images, preprocessed and augmented for better accuracy. The application uses OpenCV for real-time image capture and preprocessing, TensorFlow and Keras for building and training the CNN model, and NumPy for numerical operations.

Features

• Real-Time Detection: Captures webcam images and predicts digits in real-time.
• Custom CNN Model: Built using TensorFlow and Keras for accurate digit classification.
• Image Preprocessing: Applies grayscale conversion, histogram equalization, and normalization for better model performance.
• Data Augmentation: Uses various augmentation techniques to improve model generalization.
• Easy Integration: Simple setup and integration for real-time digit detection.

Dataset

The dataset used contains tweets labeled as either positive or negative. The dataset is preprocessed to remove noise and prepare it for training the machine learning model.

Training the Model

• Prepare your dataset: Organize your digit images into separate folders (0-9) under a main directory. Update the path variable in your script to point to this dataset directory.
• Train the model: Run the model_training.ipynb script to train the CNN model:
• This script will preprocess the images, augment the data, and train the model using TensorFlow and Keras.
• Save the model: After training, the model will be saved as model_trained.p. You can modify the script to change the model saving path.

Testing with Webcam

• Run the real-time detection: Use the main.py script to start the webcam and test the real-time digit detection:

   python main.py

• Ensure that your webcam is properly connected and accessible. The application will display the processed image and predict the digit in real-time.
• Quit the application: Press q to exit the application.

Dependencies

• Python 3.x
• OpenCV
• NumPy
• TensorFlow
• Keras
• Matplotlib
• Pickle
• Scikit-learn.

Contributing

Contributions are welcome! Please fork this repository and submit a pull request for any improvements or bug fixes.