BTAS 2015 Tutorial
Sponsored by Rank One Computing.
Have you ever been frustrated by the time it takes to design, evaluate, and
deploy a biometric algorithm? Us too!
Over the past five years we’ve been working on, and benefiting from, a
framework that streamlines the algorithm development process. We call this
framework OpenBR, and the goal of this tutorial is to turn you into a believer!
The main selling points of OpenBR include:
- A high-level grammar for composing algorithms and tuning parameters.
- A command line interface for algorithm training, evaluation, and deployment.
- A plugin framework for introducing new algorithm components.
- Hundreds of existing plugins for common algorithms (e.g., LBP, HOG,
LDA, PCA, SVM, random sampling, illumination normalization, face detection,
and landmark detection).
- A variety of standardized input and output formats.
- Automatic model file serialization and de-serialization.
- Automatic multi-threaded training, enrollment and comparison.
- Wrapper plugins for other popular open source and commercial libraries.
- Off-the-shelf algorithms for face recognition and gender/age estimation.
- Command line PDF plot generation for ROC, DET, CMC, etc. figures.
- A C++ API for custom application development.
- A library of GUI widgets for presenting OpenBR data.
- Support for Linux, OS X, and Windows operating systems.
- Daily activity on our repository and mailing list.
Computer vision scientists and software engineers interested
in a technical overview of the OpenBR project.
Prerequisites for participants
Some familiarity with computer vision, the
C++ programming language, and the Unix shell is recommended.
30-minute modules covered by different members of the OpenBR core
development team, including:
- Plugin Infrastructure - Josh Klontz, Rank One Computing
- Detection - Jordan Cheney, Noblis
- Representation - Brendan Klare, Rank One Computing
- Comparison - Josh Klontz, Rank One Computing
- Evaluation - Ben Klein, Noblis
- Application Development - Scott Klum, Rank One Computing
- IARPA Janus API & Future Directions - Josh Klontz, Rank One Computing
- J. Klontz, B. Klare, S. Klum, A. Jain, M. Burge. “Open Source Biometric
Recognition”, Proceedings of the IEEE Conference on Biometrics: Theory,
Applications and Systems (BTAS), 2013.
- Josh is a computer vision software engineer previously at MITRE,
Michigan State University and Noblis, and currently a co-founder of Rank One Computing. He is the original
author and lead developer of the OpenBR project. His professional interests
include face recognition algorithm prototyping, large-scale image
retrieval, and open source software engineering. Josh is also the author of
Likely, an open source domain specific programming language for image
- Brendan is a co-founder of Rank One Computing. Prior to this, Brendan founded the Visual Analytics Lab at Noblis.
Brendan has published over 25 peer reviewed papers on the topic of automated face recognition.
Brendan’s professional interests are in accurate algorithm design,
and developing novel applications of face recognition technology. Brendan
received his Ph.D. in Computer Science from Michigan State University.
- Scott is a computer vision research engineer is a co-founder of Rank One Computing.
He has shipped multiple biometric software
systems, including the FaceSketchID System which was licensed to Safran
Morpho in 2014. Scott has a passion for algorithm development and software
engineering with an emphasis on creating technologies that are easily
disseminated and integrated. He is the author of multiple publications,
the majority of which relate to heterogeneous face recognition.
- Ben is an engineer at Noblis. He primarily works on developing and improving
frameworks for face detection and recognition evaluation. Ben
helped to implement the Janus API in the OpenBR project.
- Jordan is a software engineer at Noblis. He recently graduated from
Swarthmore College with degrees in Computer Science and Engineering.
Currently he is focused on face detection and expanding the OpenBR