This is a handwritten digit recognizer which uses a Convolutional Neural Network model trained on the popular MNIST dataset to recognize and classify the digits being drawn on the game window that opens when you run the project.
- Languages : Python
- Libraries : Pygame, Matplotlib, Keras, NumPy, OpenCV
- Technology : Deep Learning - CNN
-
Install python-3.8 (64 bit) or above from 'https://www.python.org/downloads/' and add the PATH to your USER and System Variables under Environment variables.
-
Install VsCode or any other IDE of your choice on your device and make sure to check the add to PATH checkbox.
-
Clone this repository on your system using
git clone <repo url>
command or download as a Zip folder. -
Open the project folder in your IDE .
-
Install the necessary libraries by running the
pip install -r requirements.txt
command on your IDE's terminal. -
Run the python file named -
DrawScreen.py
-
A game window will open and you can draw digits from 0-9 for the model to recognize.
Note : You can make changes in the Model_code.py
file to modify the CNN model and then re-train it.
Source of reference --> (https://youtu.be/L2cAjgc1-bo)