IEEE Fourth International Conference on
Biometrics: Theory, Applications and Systems
The area of biometrics research has developed into a mature field with
important theoretical results, working systems, and market demand.
I general, however, the promise of biometric systems that are ubiquitous,
reliable, adaptable, robust, unobtrusive, and easy to deploy is still far
away. While good progress is being made in many areas, including
exploring new modalities, multi-modal fusion, performance evaluation, and
more useful data collection and sharing, it is useful to consider how these
efforts may fit together with current and future needs, and how progress
can be spurred to meet these needs. In this talk we will consider a set of
grand challenges for the community and discuss the possible implications for
directions in biometric research, funding, and industry.
Matt is a Professor in the Computer Science Department and Chair of the Media
Arts and Technology Graduate Program at the University of California,
Santa Barbara. His PhD thesis on Interactive Time Vision: Face Recognition
as a Visual Behavior led to one of the seminal papers on biometrics:
M. Turk and A. Pentland, "Eigenfaces for recognition," Journal of Cognitive
Neuroscience, 1991. His primary research interests are mostly concerned with
using computer vision as an input modality. That means using cameras
(and other sensors) to perceive relevant information about people -
e.g., identity, facial expression, body movement, gestures - and then using
this information to improve the perceptual interface between humans and
computers. He will chair the ninth IEEE conference on Automatic Face and Gesture Recognition in 2011.
Dr. Samy Bengio -
Large-Scale Image Annotation: Learning to Rank with Joint Word-Image Embeddings.
(pdf of slides).
Image annotation datasets are becoming larger and larger, with tens of millions of images and
tens of thousands of possible annotations.
We propose a strongly performing method that scales to such datasets by simultaneously
learning to optimize precision at k of the ranked list of annotations for a given
image and learning a low-dimensional joint embedding space for both the images and
annotations. Our method both outperforms several baseline methods and, in comparison to them,
provides a scalable architecture in terms of memory consumption and prediction time.
We also demonstrate how our method learns an interpretable model, where annotations with
alternate spellings or even languages are close in the embedding space.
Hence, even when our model does not predict the exact annotation given by a human
labeler, it often predicts similar annotations, a fact that we try to quantify by measuring
the newly-introduced "sibling" precision method, where our method also obtains
Samy Bengio is well known for his research in machine learning
and his current research interests are in large scale large margin
machine learning, mainly applied to images and sounds.
He has been a research scientist at Google since early 2007.
Before that, he was a senior researcher in statistical machine
learning at the IDIAP Research Institute in Switzerland, where he
supervised PhD students and postdoctoral fellows working on many
areas of machine learning such as support vector machines, time
series prediction, mixture models, large scale problems, speech
recognition, multi channel and asynchronous sequence processing,
multi-modal person authentication, brain computer interfaces,
text mining, and many more. He is Associate Editor of the
Journal of Computational Statistics, has been general chair of the
Workshops on Machine Learning for Multimodal Interactions
(MLMI'2004, 2005 and 2006), program chair of the IEEE
Workshop on Neural Networks for Signal Processing (NNSP'2002),
and on the program committee of several international
conferences such as NIPS and ICML.
Abstract. Confirmation bias is a tendency to interpret information in a way that confirms prior
expectations. Recently it has been demonstrated that judgements by fingerprint
experts can be influenced by contextual information. Biometric and forensic
identification ultimately relies on human judgement. In this talk identification of
facial images from CCTV by the courts will be considered. Human error in matching
facial identity, the efficacy of facial comparison methods used by expert witnesses,
and the potential influence of confirmation bias will be discussed.
Tim Valentine is a Professor of Psychology at Goldsmiths, University of London. He
obtained his PhD from the University of Nottingham in 1986. After doctoral research
at the MRC Applied Psychology Unit, Cambridge, he held lectureships at the
Universities of Manchester and Durham before being appointed to a Chair at
Goldsmiths in 1997. He has over 25 years research experience in human face
recognition and eyewitness identification. Recent projects include work on police
identification procedures, identification from CCTV and evaluation of a new method
for generating facial composites. He has provided advice to the courts on
identification issues in many criminal cases, including the appeal by Abdelbaset al
Megrahi against his conviction for the Lockerbie bomb.
Dr. Jonathon Phillips is a leading researcher in the fields
of computer vision, biometrics, face recognition, and
human identification. He is at the National Institute of
Standards and Technology (NIST), where he is
managing the Multiple Biometrics Evaluation 2010 and
the Multiple Biometrics Grand Challenge. His previous
efforts include the Iris Challenge Evaluations (ICE), the
Face Recognition Vendor Test (FRVT) 2006 and the
Face Recognition Grand Challenge and FERET. From
2000-2004, Dr. Phillips was assigned to the Defense
Advanced Projects Agency (DARPA) as program
manager for the Human Identification at a Distance
program. He was test director for the FRVT 2002. For his
work on the FRVT 2002 he was awarded the Dept. of
Commerce Gold Medal. His work has been reported in
print media of record including the New York Times and
the Economist. He has appeared on NPR's
ScienceFriday. Prior to joining NIST, he was at the US
Army Research Laboratory. He received his Ph.D. in
operations research from Rutgers University. From 2004
2008 he was an Associate Editor for the IEEE Trans. on
Pattern Analysis and Machine Intelligence and guest
editor of an issue of the Proceedings of the IEEE on
biometrics. In an Essential Science Indicators analysis of
face recognition publication over the past decade,
Jonathon's work ranks at #2 by total citations and #1 by
cites per paper. He is a Fellow of the IEEE and the