/connoisseur

Provenance Analysis in Paintings

Primary LanguageJupyter NotebookApache License 2.0Apache-2.0

Connoisseur

Battle of Grunwald “Battle of Grunwald”, Jan Matejko, 1878. From Public Domain.
Available at: wikiart.org/en/jan-matejko/battle-of-grunwald-1878.

This repository holds the experiments on Machine Learning applied to the analysis of painting provenance.
Check the INSTALL.md file for instructions on how to prepare your environment to run connoisseur.
Results can be seen at REPORTS.md.

Associated Publications

  • David L.O., Pedrini H., Dias Z., Rocha A. (2021) Authentication of Vincent van Gogh’s Work. In: Computer Analysis of Images and Patterns (CAIP). Lecture Notes in Computer Science, vol 13053. Springer, Cham.
  • L. David, H. Pedrini, Z. Dias and A. Rocha. "Connoisseur: Provenance Analysis in Paintings," IEEE Symposium Series on Computational Intelligence (SSCI), 2021.

Running Experiments

After entering the virtual environment or initiating the docker container, experiments can be found at the /connoisseur/experiments folder. An execution example follows:

cd ./experiments/
python 1-extract-patches.py with batch_size=256 image_shape=[299,299,3] \
                                 dataset_name='VanGogh' \
                                 data_dir="./datasets/vangogh" \
                                 saving_directory="./datasets/vangogh/random_299/" \
                                 valid_size=.25

This experiment will download, extract and prepare van Gogh's dataset into the data_dir directory. Finally, it will extract patches from all samples and save them in saving_directory.

Each experiment is wrapped by sacred package, capable of monitoring an experiment and logging its progression to a database or file. To do so, use the m or F parameter:

python 1-extract-patches.py -m 107.0.0.1:27017:experiments  # Requires MongoDB
python 1-extract-patches.py -F ./extract-patches/