Vikriti-ID: Synthetic Fingerprint Generator

Overview

Vikriti-ID is a synthetic fingerprint generator developed to address the challenges in fingerprint recognition research due to the limited availability of extensive and publicly available fingerprint databases. Existing databases often lack a sufficient number of identities and fingerprint impressions, which hinders progress in areas such as fingerprint-based access control.

Key Features:

• Large Database: Vikriti-ID generated a database containing 500,000 unique fingerprints, each with 10 associated impressions.
• Performance Metrics: The generated data achieved an Equal Error Rate (EER) of 0.16% and an Area Under the Curve (AUC) of 0.89%.
• Deep Learning Integration: A deep neural network, inspired by [13], was trained on both Vikriti-ID generated data as well as publicly available data, demonstrating the usability and effectiveness of the synthetic data.

Project Structure

VIKRIT-ID/
├── Code/
│ ├── Loss/
│ │ └── perceptual_loss.py
│ ├── models/
│ │ ├── composite_model.py
│ │ ├── discriminator.py
│ │ └── generator.py
│ ├── train_dataset_VAE.py
│ ├── train.py
│ ├── utils.py
│ └── VAE.ipynb
├── AtoB_generated_plot_000001.png
├── decoder.h5
├── encoder.h5
├── processed_data.npz
└── requirements.txt

• Loss/: Contains the loss function implementations.
  • perceptual_loss.py: Perceptual loss function used in model training.
• models/: Contains model architecture files.
  • composite_model.py: Composite model combining different sub-models.
  • discriminator.py: Discriminator model implementation.
  • generator.py: Generator model implementation.
• train_dataset_VAE.py: Script for training the VAE on the dataset.
• train.py: Main training script.
• utils.py: Utility functions used across the project.
• VAE.ipynb: Jupyter notebook for training and experimenting with the VAE model.
• AtoB_generated_plot_000001.png: Example plot generated during training.
• decoder.h5: Pretrained decoder weights.
• encoder.h5: Pretrained encoder weights.
• processed_data.npz: Processed dataset used for training.
• requirements.txt: Python dependencies for the project.

Citation

If you use Vikriti-ID in your research, please cite our paper:

Link to the paper

Contact

For any inquiries, please reach out to [Vansh Singh(vansh3002singh@gmail.com)] and [Aditya Sinha(adityasinha6078@gmail.com)].