Digit-Predictor

Digit-Predictor is a command-line, convolutional neural network that is able to classify handwritten digits.

The MNIST digits classification dataset was used for training; the dataset consists of 60,000 28x28 grayscale images of 10 digits, along with a test set of 10,000 images for validation during training.

Installation

The source code for Digit-Predictor can then be obtained using the git clone command with the public repository's URL as the target, to make a clone or copy of the repository in a new directory, at another location.

git clone https://github.com/bhavanvir/Digit-Predictor

Change your directory to the root of the project.

cd Digit-Predictor

Several dependencies are required to run Digit-Predictor that are not included in the Python Standard Library. It is imperative that these modules are installed and functional beforehand.

Using the package manager pip to install all external modules:

pip install -r requirements.txt

Usage

To use the built-in Digit-Predictor test harness, have the digit wanting to be predicted successfully in the name of the file. For example, if the image is of digit 3, the file name should be test_3.jpg or something similar.

If the mnist_model.h5 file is not present in the root directory, the Digit-Predictor application will first run a training session to generate the model; the default epoch length is defined as 50, but this can be changed by altering the epochs parameter within the new_model function.

model.fit(datagen.flow(x_train, y_train, batch_size=64), epochs=50, validation_data=(x_test, y_test), steps_per_epoch=x_train.shape[0] // 64, callbacks=[early_stopping_monitor])

Running the Digit-Predictor application using python3 main.py in any terminal will prompt the user with two branching options: MNIST to use the MNIST data, External to use your own external data, or Drawn to draw your own data.

1. If Upload files is selected, the user will be prompted with a Y or N option to view their processed input image, or to recieve a command-line prediction for each file located in the input directory. In either case, a prediction summary will be output to the terminal.
2. If Paint digits is selected, the user will be prompted with a paint canvas that can be used to draw a digit, then when they are satisfied with their drawing, they can select File → Predict or use the keyboard short-cut P, to recieve a prediction.

Users can also select either Y or N during application start-up, to view the confusion matrix associated with the mnist_model.h5 file; the confusion matrix serves as a visual representation of the accuracy of the model.

License

MIT