A pure Python desktop application that allows users to draw multiple digits on a canvas and predicts them using a Convolutional Neural Network (CNN) trained on the MNIST dataset. The application can detect, segment, and predict multiple digits drawn side-by-side from left to right.
- Interactive GUI: Built with Tkinter for drawing digits seamlessly.
- Multi-Digit Segmentation: Uses OpenCV to find contours and separate multiple digits drawn on the same canvas.
- Deep Learning Model: Utilizes a custom CNN built with TensorFlow/Keras, trained on the MNIST dataset to achieve high accuracy.
-
Real-time Processing: Automatically resizes, pads, and normalizes the drawn digits to match the model's expected input (
$28 \times 28$ pixels).
Make sure you have Python 3.x installed. Install the required dependencies using:
pip install -r requirements.txt