ChiMo dataset

Description

ChiMo dataset is a dataset generated using the images contained in the Chicago Face Database (CFD). This dataset contains more than 24000 morphed images.

We suggest using this dataset for the evaluation of new Single Image Morphing Attack Detectors (S-MAD)

Realization

ChiMo dataset has been generated using the images with neutral expression of the CFD dataset. For each subject, five other subjects of the same ethnicity and gender have been selected for morphing. The average face verification scores of three commercial SDKs (VeriLook, Cognitec and Innovatrics) have been used to select the most similar subjects for each individual. Then, two morphing factors (0.3 and 0.5) and three different morphing algorithms are applied for each subject pair, thus resulting in 24390 morphed images (8310 for each algorithm). Finally, two versions of this dataset are created: the first one contains digital images, while in the second we applied a compression procedure obtaining images with a maximum size of 15 kB.

Main Features:

  • 24390 morphed images (8310 for each algorithm)
  • Variety in appearances and hairstyles.
  • Two different morphing factors (0.3 and 0.5)

Release

The paper is publicly accessible. To download the data, please visit the website https://www.chicagofaces.org/resources/
We thank the authors of the original Chicago Face Database (CFD).

Reference

If you use this dataset please cite the following papers:

@article{borghi2023revelio,
  title={Revelio: a Modular and Effective Framework for Reproducible Training and Evaluation of Morphing Attack Detectors},
  author={Borghi, Guido and Di Domenico, Nicol{\`o} and Franco, Annalisa and Ferrara, Matteo and Maltoni, Davide},
  journal={IEEE Access},
  year={2023},
  publisher={IEEE}
}