October 17 at ECCV, 2008


There is rapidly increasing interest in the field of face recognition in image and videos; while there has been a great deal of progress in the past 10 years, much of the work is restricted to constrained settings in which one or more of the many variables that affect appearance, such as lighting, pose, or facial expression, has been controlled. We believe that focusing specifically on ‘real-life’ data sets will foster the development of new and more general techniques, and will ultimately result in more flexible face recognition systems. This is a new and interesting emergent direction, and we will make this workshop the right place for discussion and for the sharing of ideas on this topic.

One aim of the workshop is to encourage collaboration between researchers who do face detection and recognition but may not be familiar with more general object recognition techniques, and those who do object recognition work but have not considered the application of their methods to the face recognition problem. It is interesting to note that one of the best known algorithms for unconstrained face recognition using random forests does not appear on the “Face Recognition Web Page”. This is probably because the algorithm is known as a generic object recognition algorithm. We want to bring together people from the “object recognition” community and the “face recognition” community, and try to understand if they are distinct only for historical reasons or if they rely on different foundations.


• Erik Learned-Miller (Primary contact)

University of Massachusetts, Amherst, USA

• Andras Ferencz
Mobileye Vision Technologies, Princeton, New Jersey, USA

• Frédéric Jurie
LEAR Group, CNRS, INRIA, France

Format and Schedule

Date: Friday, October 17.
Time: 2pm-5:30pm

SESSION 1: 2:00-3:30

2:00-2:20 -- Introduction to Faces in Real-Life Images
Erik Learned-Miller

2:20-2:40 -- Descriptor Based Methods in the Wild
Lior Wolf, Tal Hassner, Yaniv Taigman

2:40-3:00 -- Face Recognition in Unconstrained Environments: A Comparative Study
Rodrigo Verschae, Javier Ruiz-del-Solar and Mauricio Correa

3:00-3:20 -- Establishing Good Benchmarks and Baselines for Face Recognition
Nicolas Pinto, James J. DiCarlo, and David D. Cox

3:20-3:25 -- LFW Results Using a Combined Nowak Plus MERL Recognizer
Gary B. Huang, Michael J. Jones, and Erik Learned-Miller

COFFEE BREAK: 3:30-4:00

SESSION 2: 4:00-5:30

4:00-4:30 -- Invited Speaker (changed): Peter Belhumeur
Peter Belhumeur, Columbia University

4:30-4:50 -- Open Discussion: Important Directions in Face Recognition Research.

4:50-4:55 -- Combined Model for Detecting, Localizing, Interpreting, and Recognizing Faces
Karlinsky et al.

4:55-5:00 --Learning Shape Metric: From Alignment to Recognition
Daniel Gill and Yaniv Ninio

5:00-5:05 -- Simultaneous Learning and Alignmennt: Multi-Instance and Multi-Pose Learning
Boris Babenko, Piotr Dollar, Zhuowen Tu, and Serge Belongie

5:05-5:10 -- A Prototype System for Unconstrained Face Verificatioin Based on Statistical Learning
Augusto Desrero, Alberto Lovato, and Francesca Odone

5:10-5:15 -- Aligning Names and Faces: Seeing the Problem in Different Ways
Phi The Pham, Marie-Francine Moens, and Tinne Tuytelaars

5:15-5:30 -- POSTER SESSIONS