Git Product home page Git Product logo

fingerflow's Introduction

FingerFlow

Python PyPI

FingerFlow is an end-to-end deep learning Python framework for fingerprint minutiae manipulation built on top of Keras - TensorFlow high-level API.

In current stable version 3.0.1 following modules are provided:

  • extractor - module responsible for extraction and classification of minutiae points from fingerprints. It is also capable of detecting fingerprint core points.
  • matcher - module responsible for matching extracted minutiae feature vectors.

GPU support

FingerFlow supports GPU acceleration on CUDA®-enabled graphic cards.

Software requirements

  • Python 3.7 or newer
  • CUDA - for TensorFlow GPU acceleration (if missing, CPU will be used for computation)

Installation

Use the package manager pip to install FingerFlow. We reccomend to use it in pip or anaconda enviroment.

Installation in anaconda enviroment:

pip install fingerflow

Modules

Extractor

Module responsible for extraction and classification of fingerprint minutiae points and also for detecting fingerprint core point. Minutiae extraction part is built using MinutiaeNet neural network architecture. Core detection part is built using YOLOv4 object detection neural network architecture.

Extractor contains 3 modules:

  • MinutiaeNet - module responsible for extracting minutiae points from fingerprint image. Using MinutiaeNet neural network architecture.
  • ClassifyNet - module responsible for classifying extraced minutiae points. Architecture based on FineNet module of MinutiaeNet.
  • CoreNet - module responsible for detecting fingerprint core point. YOLOv4 neural network architecture is used.

Neural networks models

API

Extractor

Class which provides all functionality for extraction of minutiae points

fingerflow.extractor.Extractor()

Arguments

  • coarse_net_path - used for setting path to pretrained model of submodule CoarseNet
  • fine_net_path - used for setting path to pretrained model of submodule FineNet
  • classify_net_path - used for setting path to pretrained model of submodule ClassifyNet
  • core_net_path - used for setting path to pretrained model of submodule CoreNet

Methods

  • extract_minutiae(image_data) - used for extracting minutiae points and detecting of fingerprint core from input image data. Method accepts input data in form of 3D matrix (e.g. output of OpenCV imread function). Method returns object with extracted and detected data in following shape:
  • minutiae - Pandas DataFrame of extracted and classified minutiae points in following form:
    • x - x coordinate of minutiae point
    • y - y coordinate of minutiae point
    • angle - direction of minutiae point rotation
    • score - minutiae point extraction confidence
    • class - type of minutiae point. In FingerFlow 1.0.0 we support following minutiae classes:
      • ending
      • bifurcation
      • fragment
      • enclosure
      • crossbar
      • other
  • core - Pandas DataFrame of deteted fingerprint cores in following form:
    • x1 - left coordinate of bounding box
    • y1 - top coordinate of bounding box
    • x2 - right coordinate of bounding box
    • y2 - bottom coordinate of bounding box
    • score - core detection confidence
    • w - width of bounding box
    • h - height of bounding box

Usage

import cv2
import numpy as np
from fingerflow.extractor import Extractor

extractor = Extractor("coarse_net", "fine_net", "classify_net", "core_net")

image = cv2.imread("some_image")

extracted_minutiae = extractor.extract_minutiae(image)

Matcher

Module responsible for matching extracted feature vectors. It is using custom Siamese neural network architecture. Input size (number of minutiae in feature vector) for matching is not fixed and is determined by precision constructor argument. Used weights need to be in the correct shape - network needs to be trained with the same precision as passed in the argument. But in general, the more minutiae points the higher precision. Our custom model is trained on 10, 14, 20, 24, 30 minutie points per input.

Alt text

Matcher contains 1 module:

  • VerifyNet - module responsible for matching feature vectors. Custom Siamese neural network architecture is used.

Neural network models

API

Matcher

Class which provides all functionality for matching feature vectors

fingerflow.matcher.Matcher()

Arguments

  • precision - input size (number of minutiae in feature vector)
  • verify_net_path - used for setting path to pretrained model of submodule VerifyNet

Methods`

  • verify(anchor, sample) - used for matching feature vectors. Method accepts input data in form of NumPy N-dimensional array, which should be in following shape: (confidence, columns). Its columns should contain the same data as Extractor minutiae output with one additional column that represents minutia distance to fingerprint core. Method returns float matching confidence in range 0-1.
  • verify_batch(pairs) - performs pairwise verification in provided list of pairs. Returns list of float matching confidences.
  • plot_model(file_path) - plot model structure into image file. file_path should represent path to desired file.

Usage

from fingerflow.matcher import Matcher

matcher = Matcher(30, "verify_net")

prediction = matcher.verify(anchor_feature_vector, sample_feature_vector)

Contributing

Pull requests are welcome. For major changes, please open an issue first to discuss what you would like to change.

License

MIT

fingerflow's People

Contributors

jakubarendac avatar

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.