This project is a simple yet effective tool that recognizes single handwritten numbers with an accuracy rate of 93%. Users can write any number on a 28x28 virtual whiteboard, and the recognizer provides a prediction within 3 seconds.
The recognizer utilizes a basic Convolutional Neural Network (CNN) trained on handwritten digit data. By analyzing 28 pixels line by line, the model predicts the number based on the patterns formed by the lines. Key libraries such as NumPy and TensorFlow are employed for data processing and model computations.