Handwritten Mathematical Equation Recognition Using CNN

This project is a web-based application that recognizes and solves handwritten mathematical equations using a Convolutional Neural Network (CNN). It provides a user-friendly interface to either upload an image of an equation or draw it directly on a canvas.

Features

• Equation Recognition: Uses a trained CNN model to recognize handwritten digits and mathematical symbols.
• Equation Solver: Solves the recognized equations and provides the result.
• Input Methods:
  • Upload Image: Upload an image file containing a handwritten equation.
  • Canvas Drawing: Draw an equation directly on the screen using the built-in canvas.
• Web Interface: Built with Flask for a simple and interactive user experience.

Tech Stack

• Backend Framework: Flask
• Deep Learning: TensorFlow / Keras
• Image Processing: OpenCV, Pillow (PIL)
• Computer Algebra: SymPy
• Data Handling: Pandas

Installation

1. Clone the repository:

   git clone  https://github.com/cepdnaclk/e20-co542-Handwritten-Mathematical-Equation-Recognition-Using-CNN.git
   cd e20-co542-Handwritten-Mathematical-Equation-Recognition-Using-CNN
   

2. Install the required dependencies:

   pip install -r requirements.txt
   

Usage

1. Run the Flask application:

   python app.py
   

2. Open your web browser and navigate to:

   http://127.0.0.1:5000/
   

3. To Use:
   • Upload: Click on “Upload Image” to select a file from your computer.
   • Canvas: Click on “Canvas” to draw an equation.
   • Click the “Predict” or “Solve” buttons to get the result.

Project Structure

• app.py: Main Flask application file defining routes and logic.
• requirements.txt: List of Python dependencies.
• train_classifier.ipynb: Jupyter Notebook used for training the CNN model.
• eqn-detect-new-model.keras: Trained model file.
• equation_calculator.py & solve_equation_file.py: Helper scripts for equation parsing and solving.
• templates/: HTML templates for the web interface.
• static/: Static files (CSS, JS, images).