Special Tutorial

The tutororial by Sandy Pentland (MIT) has been cancelled.

Tutorial: Introduction to PyVision and CSU Face Recognition Baseline Algorithm

Date: Monday, January 9, 2012

Presenters: David Bolme and Steve O’Hara – Colorado State University

As part of continuing evaluation work CSU is releasing new open-source face recognition algorithms. There are many reasons for releasing these algorithms. First, these algorithms provide a standard to assess relative difficulty of new data sets. Second, they provide a means of characterizing the relative difficulty within different parts of major evaluations. Third, because their internal mechanisms are based on well known methods, the algorithm development community can benefit by observing their behavior relative to other algorithms. The baseline system provides a complete end-to-end automatic face recognition framework including Viola and Jones face detection, ASEF eye detection, normalization, PCA and LDA dimensionally reduction, cohort score normalization, and support for Biometrics Evaluation Environment formats and methods. The system serves as both a benchmark for performance as well as a framework for new algorithm development.

This hands on, half day tutorial will introduce the major components of the evaluation system. Session 1 which is intended for those new to Python will discuss installation of Python and the evaluation system and will also give a brief introduction to Python and SciPy as a scientific computing language similar to Matlab. Session 2 will provide an introduction to the PyVision Computer Vision Toolkit which provides many components of the baseline system. Session 3 will present a summary the CSU baseline algorithms and how to use the system to conduct an evaluation. Session 4 will allow participants to run the system on a small dataset and get hands on experience with the evaluation framework. Participants are encouraged to download the software before the tutorial and follow along with the demos.

Visit the Tutorial’s Site

Comments are closed.