Robust and Low-Rank Representation for Fast Face Identification with Occlusions

by Michael Iliadis, Haohong Wang, Rafael Molina, Aggelos K. Katsaggelos, published on Transactions on Image Processing, May 2017 and ArXiv.

Introduction

The code provides MATLAB implementation of the F-LR-IRNNLS and the F-IRNNLS algorithm which is a fast version of the RRC Regularized robust coding for face recognition algorithm (with non-negative representation coefficients - but it's easy to adapt it to L2 or L1 coeff.)

Usage

Download MATLAB code

Clone the repository:

$ git clone https://github.com/miliadis/FIRC
$ cd ~/FIRC 

Data

• By default the experiments reproduce the FIRC results on YaleB dataset with 60% occlusion.
• Data of YaleB are provided to reproduce the results of the paper. For more data please reach out.

Testing

1. Add to the path the entire "FIRC" folder

2. Run "Main.m"

   Main.m

3. Results will be printed for each query face.

Run F-LR-IRNNLS vs. F-IRNNLS

1. If you want to run the F-LR-IRNNLS classifier for occlusions set:

   fr.alg = 'f-lr-irc'

2. If you want to run the F-IRNNLS which is the fast version of the RRC algorithm from the "Regularized robust coding for face recognition" paper set fr.alg = 'f-irc'.

   fr.alg = 'f-irc'

   Make sure that the settings of the weight function are similar to the ones in RRC.

Citation

If FIRC is useful for your research, please consider citing:

@article{Iliadis2017, 
author={M. Iliadis and H. Wang and R. Molina and A. K. Katsaggelos}, 
journal={IEEE Transactions on Image Processing}, 
title={Robust and Low-Rank Representation for Fast Face Identification With Occlusions}, 
year={2017}, 
volume={26}, 
number={5}, 
pages={2203-2218},
month={May}}

Questions

Please contact 'miliad@u.northwestern.edu'