Deep Auto-Encoding and Biohashing for Secure Finger Vein Recognition

This package is part of the signal-processing and machine learning toolbox Bob. It contains the source code to reproduce the following paper:

@inproceedings{ICASSP2021_DAE_SecureFVR,
  title={Deep Auto-Encoding and Biohashing for Secure Finger Vein Recognition},
  author={Shahreza, Hatef Otroshi and Marcel, S{\'e}bastien},
  booktitle={ICASSP 2021-2021 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
  pages={2585--2589},
  year={2021},
  organization={IEEE}
}

If you use this package and/or its results, please cite both the package and the corresponding paper. Also, please ensure that you include the two original Bob references in your citations.

NOTE: In the instructions that follow, the use of < ... > inside file paths indicates that you must replace the part enclosed within < and > with the appropriate value. For example <path_to_your_working_directory> indicates that you should replace this part with the full path to the directory inside which you are running this code.

Installation

The experiments can only be executed on a Linux 64-bit machine. Install conda and run the steps below::

$ git clone https://gitlab.idiap.ch/bob/bob.paper.icassp2021_deepae_biohashing_securefvr.git
$ cd bob.paper.icassp2021_deepae_biohashing_securefvr
$ conda create -n ICASSP2021_DAE_SecureFVR --file package-list.txt
$ conda activate ICASSP2021_DAE_SecureFVR

Downloading the Databases

The next thing you must do is to download the UTFVP fingervein database used in our experiments. Please refer to the this link to download this database.

Take note of the directory in which you have stored the downloaded database. Then, create a textfile named .bob_bio_databases.txt and store it in your home directory. Inside this textfile, insert the following line:

[YOUR_UTFVP_DIRECTORY] = <path_to_utfvp_database>

Make sure you replace <path_to_utfvp_database> with the path to your downloaded UTFVP database.

Running the Experiments

You are now ready to run the experiments to reproduce the results in the paper. To do this, follow steps 1-5:

Step 1: Preparing data for training (Data Augmentation)

In order to prepare the data for training the neural network you need to run data.py in data directory:

$ cd data
$ python data.py

NOTE: In data directory, there are two Jupyter notebooks to work more with the dataset:

• Data-Augmentation.ipynb: This notebook contains the code for data augmentation as well as code for visualization of augemnted data.
• Sample_Image.ipynb: This notebook contains the code to generate images used in figure 1 of the paper.

Step 2: Training Deep Auto-Encoder

Now that the data is ready, we can train our DNN. For this end, run the following code:

$ cd models/img
$ python train.py

Step 3: Executing different finger vein recognition methods

We can execute different finger vein recognition methods including WLD, WLD+Biohash, MC, MC+Biohash, RLT, RLT+Biohash and our trained AE+Biohash [the proposed method] by runnung the following script:

$ execute_fvr.sh

Step 4: Generating Figure 3 and Table 1

In order to evaluate and compare the FVR models, you can run the following script:

$ cd evaluation/compare_methods
$ ./eval.sh

This generates two .pdf files and two .txt files which are used in Figure 3 and Table 1, respectively.

Step 6: Generating Table 2

In order to evaluate and compare the execution time of different FVR models as presented in Table 2 of the paper, you can use the Jupyter notebooks in evaluation/time directory.

Contact

For questions or reporting issues to this software package, contact our development mailing list.