Match photos in newspapers to photos on a reel

Install the required packages (in a virtual environment) using the following commands

python -m pip install -r requirements.txt
python main.py

Usage

The script main.py accepts two command line arguments:

• input_folder, pointing to a folder containing:
  • A subfolder newspaper, containing:
    • .jpg images of newspaper scans
    • .xml files specifying image coordinates
  • A subfolder photo, containing:
    • .jpg images of (cropped) photos
  • A config.json, specifying the configuration of the data
• output_folder, pointing to a folder for the output, default is output. If this folder does not yet exist, it will be created. If it does exist, existing results will be overwritten. It will be populated with:
  • for every photograph from the list find N-best matches from the illustrations in newspapers
  • verify whether match is 'true/false' (postprocessing)

See the examples in the tests folder for more details.

Tests

There is a single test that performs the whole procedure for two different archives, using different configuration.