With this Dockerfile you can create a Docker image within which you can run Ipython Notebooks that train and test various neural network models to classify Fashion-MNIST data.

This Dockerfile simply clones a Github repository which has the code.

See also cnn2_model_out folder for the output of the best naural network model I got so far (Image augmentation + CNN with dropout). You can reproduce the results through the code. After applying the couple of steps below, you can see code repository's Readme for more details about the code.

Follow the steps below to run the app:

• Docker

• Clone the repository into your local. git clone https://github.com/berfubuyukoz/fashion_mnist_case.git

• Make sure you are in the project directory, fashion_mnist_case

• In order to build the image run the following command (mind the dot at the end of the command): This will take 1-2 minutes.

  docker build --tag=fmnist .
  

• Run docker image on port 8888

  docker run -d -v /$(pwd)/:/home/jovyan/work -p 8888:8888 fmnist start-notebook.sh --NotebookApp.token=''  
  

• Open browser, write localhost:8888/?token=

where is the token appeared on terminal.