Products, open source tools, and research lines: the things I build along the way.
A corporate AI platform that connects documents, data and processes and turns them into operational knowledge and traceable decisions: an assistant over millions of documents, document analysis and generation, BI, and specialized agents.
A Python package for evaluating biometric and binary classification systems: EER and other operating points, ROC/DET curves for verification, and CMC curves for identification.
NumPy/SciPy implementations of state-of-the-art global and local image thresholding algorithms.
Can commodity document scanners replace dedicated hardware for capturing palmprints? A feasibility study of minutiae extraction, a scanner prototype, and GAN-based style transfer between acquisition devices.
A technical review and comparative study of state-of-the-art deep learning techniques for extracting fingerprint minutiae.
Spatial-domain algorithms and neural networks for extracting palmprint principal lines from high-resolution images, plus alignment for line correlation in forensic use.
Using singular points to cut the cost of high-resolution palmprint matching.
Frequency-domain analysis and contextual filter design for enhancing latent fingerprints and palmprints.
Automatic digitization of fingerprint cards: texture analysis to classify card models and segmentation of rolled and slap impressions.
Neural networks that determine whether a palmprint comes from a left or a right hand.
A compact speaker representation in a Bernoulli space, with channel variability compensation and SVM-based verification.
An image-processing and neural network pipeline for recognizing handwritten digits.
