Realistic Fingerprint Generator

Tensorflow implementation of realistic fingerprint generator based on DCGAN Deep Convolutional Generative Adversarial Networks.

Prerequisites

• Python 3.6+
• Tensorflow 1.12.0
• SciPy
• pillow

For training:

• Download NIST Special Database 14.
• Move to "./data" folder

For testing:

• Download Pre trained model
• Extract checkpoint folder to the main root (under "-Realistic-Fingerprint-Generator").

Usage

Generate fingerprint with pre-trained model (NIST14):

$ python main.py --input_height=650 --output_height=650 --test

To train a model with NIST14 dataset:

$ python main.py --dataset nist14 --input_height=650 --output_height=650 --train --crop=True

Or, you can use your own dataset (without central crop) by:

$ mkdir data/DATASET_NAME
... add images to data/DATASET_NAME ...
$ python main.py --dataset DATASET_NAME --train
$ python main.py --dataset DATASET_NAME
$ # example
$ python main.py --dataset=Fimages --input_fname_pattern="*.tif" --train

Evaluation

Use state-of-the-art fingerprint feature extractor fingerNet to extract some statistics from real and generated images

Data set	Mean number of minute	STD number of minute	Mean orientation of minute	STD orientation of minute
Nist14	123.39	30.63	3.30	1.85
Generated	123.87	30.53	3.29	1.88

Related works

• DCGAN-tensorflow
• fingerNet

Author

Rafael Bouzaglo / [@rafaelbou]