Handwriting Digit Recognition

This project enables the recognition of handwritten digits using TensorFlow and Tkinter libraries with mnist dataset. After drawing a digit on the Tkinter interface, the TensorFlow model is used to predict the drawn digit.

I used CNNs (Convolutional Neural Networks) and data augmentation techniques to get high val-accuracy result.

Preview

Hand.Writing.Digit.Recognition.mp4

Model

I used CNN (Convolutional Neural Networks) and data augmentation techniques in my model

Model Accuracy

Final training loss: 0.0444

Final training accuracy: 0.9858

Final validation loss: 0.0182

Final validation accuracy: 0.9948

Technologies Used

• Python 3: The project is developed using Python programming language.
• Pillow (PIL): Utilized for capturing and processing images.
• TensorFlow: Used for training the data, loading pre-trained models and making predictions.
• NumPy: Employed for array manipulation and normalization of input data.
• Tkinter: Utilized for creating the user interface (canvas).
• Google Colab: Used for fast model training with GPUs.