Paper Fingerprinting using Deep Learning

Jul 10, 2021
PDF
Finding Unique Fingerprints in Microscopic Paper Textures

  • Built a unique image embedding generation pipeline for document fraud detection using Beta-Variational Autoencoders.
  • Employed the use of Shearlet Transforms, Variational Autoencoders and Generative Networks for robust Fingerprint generation using a specialized Perceptual Loss function.
  • Generated a custom dataset of 55,000 Paper Texture Images for efficient model training.
  • Benchmarked performance of Gabor Filters & Local Binary Patterns for Paper Fingerprint Robustness using Euclidean Distance and Bhattacharyya Distance, on the custom dataset.