/fingerprint-synthesis

Fingerprint Synthesis: Evaluating Fingerprint Search at Scale

Primary LanguagePython

Synthesizing Fingerprint Images

The code is an implementation of the paper "Fingerprint Synthesis: Evaluating Fingerprint Search at Scale, in ICB 2018" by K. Cao and A. K. Jain.

Further work to synthesize a dataset of diverse 100 million fingerprint images can be found in this paper "Fingerprint Synthesis: Search with 100 Million Prints" by V. Mistry, J. J. Engelsma and A. K. Jain.

Dependencies

  • Python 3
  • Tensorflow
  • Tensorpack 0.5
  • Numpy
  • Scipy

Pretrained Model

Please download the model here. Please extract the zipped file into model directory in root.

Usage

The bash script 'Generate_Fingerprints.sh' will run the code for synthesizing fingerprint images. Running the bash script can be done simply by:

bash Generate_Fingerprints.sh

The script has 3 flags for specifying the following:

  • The model to load (--load).
  • The directory to store the generated fingerprint images (--sample_dir).
  • The number of fingerprints to synthesize (--num_images).

You can change the values of these flags in the bash script.

Citations

Please cite the following paper:

"Fingerprint Synthesis: Evaluating Fingerprint Search at Scale, in ICB 2018" by Kai Cao and Anil K. Jain.

@inproceedings{8411200,
author={K. {Cao} and A. {Jain}},
booktitle={2018 International Conference on Biometrics (ICB)},
title={Fingerprint Synthesis: Evaluating Fingerprint Search at Scale},
year={2018}}