CNN-project

Goal:

The goal of this project is to train a convolutional neural network to correctly classify handwritten digits.

Data:

The dataset used for this project will come from MNIST (http://yann.lecun.com/exdb/mnist/) which provides a dataset with training images and their labels, as well as a dataset with test images and their correct labels.

Algorithm:

We intend to use a convolutional neural network to solve this problem.

Tools:

We will implement this project in Python using the TensorFlow library’s implementation of a CNN.

Using Docker:

Build and Run the Container

docker build -t cnn-project .
docker run -v "$(pwd)":/cnn_project -it cnn-project

Now you're in the Docker container!
Run "python test.py" to verify everything is working.
Type "exit" to leave container.

You can also use wsl and docker desktop to get the container going.
Type "code ." to open vscode in the current directory.

Compile and Test a CNN Model

python your_model_file.py
python test_cnn.py your_cnn_model.model

Group Members:

• Eric Galvan
• Kendal Hasek
• Nicholas Hibbard
• Chloe Koschnick
• Dylan Wright

Citations:

This project was inspired by a video tutorial entitled Neural Network Python Project - Handwritten Digit Recognition, created by NeuralNine. You can watch the video here.