DeepJanus-MNIST

Test Input Generator for MNIST - Getting Started

General Information

This folder contains the application of the DeepJanus approach to the handwritten digit classification problem. The following instructions allow to rapidly run DeepJanus, without configuring your environment from scratch.

NOTE: If you want to configure your machine to run DeepJanus-MNIST, please read our detailed installation guide

Step 1: Configure the environment

Pull our pre-configured Docker image for DeepJanus-MNIST:

docker pull p1ndsvin/ubuntu:dj

Run the image by typing in the terminal the following commands:

docker run -it --rm p1ndsvin/ubuntu:dj
cd venvs
. .djvenv/bin/activate
cd ..

Step 2: Run DeepJanus

Use the following commands to start a rapid run of DeepJanus-MNIST:

cd deepjanus/DeepJanus-MNIST
python main.py

Usage

Input

• A trained model in h5 format. The default one is in the folder models;
• A list of seeds used for the input generation. The default list is in the folder original_dataset;
• config.py containing the configuration of the tool selected by the user.

Output

When the run is finished, the tool produces a folder named run_x (where x is the timestamp of the run) located in the folder runs. The output folder contains the following outputs:

• config.json reporting the configuration of the tool;
• stats.csv containing the final report of the run;
• the folder archive containing the generated inputs (both in npy array and image format).

Troubleshooting

• If tensorflow cannot be installed successfully, try to upgrade the pip version. Tensorflow cannot be installed by old versions of pip. We recommend the pip version 20.1.1.
• If the import of cairo, potrace or other modules fails, check that the correct version is installed. The correct version is reported in the file requirements.txt. The version of a module can be checked with the following command:

$ pip show modulename | grep Version

To fix the problem and install a specific version, use the following command:

$ pip install 'modulename==moduleversion' --force-reinstall