/02238-Biometric-Systems

Personal Code repository for June 2018 Summer course (02238 Biometric Systems)

Primary LanguagePython

Finger Knuckle Print (FKP)

02238-Biometric-Systems

Assignment Link

Paper formalities

Data

The dataset, THU-FVFDT21 contains Region-of-Interest (ROI) images of finger dorsal texture from 610 different subjects. A subject's sample consists of two normalized images of the same ROI - taken separately and normalized to 200x100 pixels. The first 390 subjects are sampled within 3-7 days. The remaining 220 subjects are sampled within approximately half a minute.1

Pre-processing

Training

Training is based loosely on Keras MNIST example2.

Acronyms:

  • FAR - False Acceptance Rate
  • GAR - Genuine Acceptance Rate
  • ROC - Receiver Operator Characteristic
  • DET - Detection Error Tradeoff

Research papers

  • Kannala3
  • Kumar4
  • Zhang5
  • Siamese CNN6
  • Siamese Neural Networks for One-shot Image Recognition7
  • Old Paper on siamese networks8

Footnotes

  1. Tsinghua University Finger Vein and Finger Dorsal Texture Database ↩ ↩2

  2. Keras MNIST Example GitHub ↩

  3. Kannala-BSIF-(2012) ↩

  4. Kumar-FK-(2009) ↩

  5. Zhang-FKP-(2009) ↩

  6. https://ethz.ch/content/dam/ethz/special-interest/baug/igp/photogrammetry-remote-sensing-dam/documents/pdf/learning-tracking-siamese.pdf ↩

  7. https://www.cs.cmu.edu/~rsalakhu/papers/oneshot1.pdf ↩

  8. http://papers.nips.cc/paper/769-signature-verification-using-a-siamese-time-delay-neural-network.pdf ↩