Handwritten Digit Recognition

A model that trains a model to recognize handwritten digits using the MNIST data set. Then it loads external files and uses the neural network to predict what digits they are.

Handling Digit Recognition Failures

If the model fails to recognize digits accurately, consider the following solutions:

1. Increase Training Epochs: More training epochs can help the model learn better. Try increasing the number of epochs in the model.fit() method.

2. Image Preprocessing: Ensure that input images are preprocessed consistently with the training images. For example, ensure images are properly normalized and resized. 3 Augment Data: To improve model robustness, you can use data augmentation techniques to generate variations of the training images, which can help the model generalize better.

3. Tune Hyperparameters: Adjust hyperparameters such as the number of neurons in the Dense layers, learning rate of the optimizer, or the choice of activation functions.

4. Check Model Architecture: Experiment with different architectures, such as adding more layers or using different types of layers (e.g., Convolutional layers).

5. Ensure Proper Image Format: Verify that the images you are using for prediction are correctly formatted and aligned with the model’s expected input dimensions and normalization.

Libraries Used:

• TensorFlow for providing the tools to build and train the model.
• MNIST Dataset for the handwritten digits dataset.

Resolve all Dependencies:

pip install -r requirements.txt

Feel free to add 28x28 pixel images into the digits directory!