Descriptor Based Methods in the Wild

Lior Wolf1    Tal Hassner2    Yaniv Taigman1,3

1. The School of Computer Science,
Tel-Aviv University, Israel

2. Computer Science Division
The Open University of Israel

3. face.com,
Tel-Aviv, Israel

 

Abstract: Recent methods for learning similarity between images have presented impressive results in the problem of pair matching (same/not-same classification) of face images. In this paper we explore how well this performance carries over to the related task of multi-option face identification, specifically on the Labeled Faces in the Wild (LFW) image set. In addition, we seek to compare the performance of similarity learning methods to descriptor based methods. We present the following results: (1) Descriptor-Based approaches that efficiently encode the appearance of each face image as a vector outperform the leading similarity based method in the task of multi-option face identification. (2) Straightforward use of Euclidean distance on the descriptor vectors performs somewhat worse than the similarity learning methods on the task of pair matching. (3) Adding a learning stage, the performance of descriptor based methods matches and exceeds that of similarity methods on the pair matching task. (4) A novel patch based descriptor we propose is able to improve the performance of the successful Local Binary Pattern (LBP) descriptor in both multi-option identification and same/not-same classification.

Reference:  Lior Wolf, Tal Hassner and Yaniv Taigman, "Descriptor Based Methods in the Wild," Faces in Real-Life Images workshop at the European Conference on Computer Vision (ECCV), Oct 2008.

Click here for the PDF (1,098kb)

Click here for the BibTex

Best result on the LFW face recognition benchmark, at the Faces in Real-Life Images workshop, 2008
Related publications:

Multiple One-Shots for Utilizing Class Label Information

The One-Shot Similarity Kernal

MATLAB Code for the Three-Patch LBP (TPLBP) and Four-Patch LBP (FPLBP) global image descriptors

Below please find MATLAB code for producing the TPLBP and FPLBP codes and global image descriptors. Note that this code was not the one used in the experiments reported in the paper, nor are the default parametrs necessarily the same as the ones we used. Please report any bugs or problems to hassner@openu.ac.il.

Type "help TPLBP" or "help FPLBP" for more information on each of these functions.

A typical usage would look something like this:

>> I = imread(...);
>> I = rgb2gray(I);
>> [descI, codeI]=FPLBP(I);
>> descI = descI(:);

descI will then contain the FPLBP global image descriptor.

Downloads:

TPLBP.m
FPLBP.m

Copyright and disclaimer:

Copyright 2008, Lior Wolf and Tal Hassner

The SOFTWARE ("FPLBP.m" and \ or  "FPLBP.m") is provided "as is", without any guarantee made as to its suitability or fitness for any particular use.  It may contain bugs, so use of this tool is at your own risk. We take no responsibility for any damage that may unintentionally be caused through its use.
Copyright: no material is allowed to be copied or used in any way without written permission of the authors.
Last update 22th of July, 2009