Image Featurization

This is the featurization step for the project "Walking through the Music and Visual Art StyleSpace via Computational Intelligence" (1).

Here we extracted the features from the penultimate layer of the network implemented by the authors Adrian Lecoutre, Benjamin Negrevergne and Florian Yger (2)(3), without performing any retraining procedures.

1. https://github.com/erikagardini/Walking-through-the-Music-and-Visual-Art-StyleSpace-via-Computational-Intelligence

2. Adrian Lecoutre, Benjamin Negrevergne and Florian Yger (2017). Recognizing Art Style Automatically in painting with deep learning. JMLR: Workshop and Conference Proceedings (80) 1–17.

3. https://github.com/bnegreve/rasta

Install python requirements

You can install python requirements with

pip3 install -r requirements.txt

Downloads files

git clone https://github.com/erikagardini/ImageFeaturization.git

Downloads the Wikipainting dataset

Download the full wikipaintings dataset (the one from WikiArt) executing the following commands. Warning: the file is about ~20GiB.

cd datasets
wget www.lamsade.dauphine.fr/~bnegrevergne/webpage/software/rasta/wikipaintings_full.tgz
tar xzvf wikipaintings_full.tgz
cd ../

Downloads the rasta models

cd models
wget www.lamsade.dauphine.fr/~bnegrevergne/webpage/software/rasta/rasta_models.tgz
tar xzvf rasta_models.tgz
cd ../

How to use the code

Step 1: Test the model

Test the model obtained during the rasta experiment.

cd ../python
python3 1_testing.py

When the testing is completed, the file img_dataset_testing.csv is saved inside the directory outputs and contains the output of the penultimate layer of the network during the testing.

Step 5: Formatting the image dataset

python3 2_format_dataset.py

Here, the dataset img_dataset_testing.csv is correctly formatted. The output is the dataset img_dataset.csv, which is used for the experiment "Walking through the Music and Visual Art StyleSpace via Computational Intelligence" (1).