########################################################################### # # Copyright 1997 by the Massachusetts Institute of Technology. All # rights reserved. # # Developed by the Agents Group and the Vision and Modeling Group # at the Media Laboratory, MIT, Cambridge, Massachusetts. # # Permission to use, copy, or modify this software and its documentation # for educational and research purposes only and without fee is hereby # granted, provided that this copyright notice and the original authors's # names appear on all copies and supporting documentation. If individual # files are separated from this distribution directory structure, this # copyright notice must be included. For any other uses of this software, # in original or modified form, including but not limited to distribution # in whole or in part, specific prior permission must be obtained from # MIT. These programs shall not be used, rewritten, or adapted as the # basis of a commercial software or hardware product without first # obtaining appropriate licenses from MIT. MIT. makes no representations # about the suitability of this software for any purpose. It is provided # "as is" without express or implied warranty. # ########################################################################### FERET face recognition software and the FERET pipeline software are available for educational and research purposes only, to those sponsoring, affiliated with, or collaborating with the Media Lab. Please do not redistribute without permission. The files in this directory are: README (this file) pipe.tar.gz (FERET pipeline software) feret.tar.gz (FERET face recognition software source and binaries) This release is intended for advanced users of the face recognition package and familiarity with the algorithm, the Unix environment as well as Matlab and C code is necessary. If there are any problems or questions, please mail them to us and we will do our best to help you out. Tony Jebara jebara@media.mit.edu FERET pipeline software ======================= The short form: mkdir pipe mv pipe.tar.gz pipe cd pipe gunzip pipe.tar.gz tar xvf pipe.tar THere is an excellent documention for the code in ./BTpipe/pipe/src/faces/THE_STORY THE_STORY is *Extremely* useful documention to have even if you are using the FERET code since it uses a lot of the same functions and discusses them more completely. There are other README and pseudo-instructions scattered about in the subdirectories where they are needed. FERET face recognition software source and binaries =================================================== The short form: mkdir feret mv feret.tar.gz feret cd feret gunzip feret.tar.gz tar xvf feret.tar The following subdirectories should be present in the feret directory you have just created: Clean/ Code/ Data/ bin/ In ./Data the following subdirectories are visible: Bayesian/ Eigenfaces/ Features/ Headsearch/ Masks/ Models/ ./Data/Eigenfaces/ff contains 125 eigenfaces trained on 2000+ images of aligned faces ./Data/Bayesian/ contains intra-class and extra-class eigenvectors in the coefficient space generated by projecting the images onto the eigenfaces in Data/Eigenfaces/ff ./Data/Bayesian/ has the following eigenvector files: extra_kl.125.bf extra_kl.bf intra_kl.200.bf extra_kl.125.new.bf intra_kl.125.bf intra_kl.bf extra_kl.200.bf intra_kl.125.new.bf The way the system worked was you projected coefficient-differences(as opposed to image-differences) on to the intra eigenvectors and then on to the extra eigenvectors and computed a likelihood ratio test. The code for this can be found in a file called neighbor.c in the ./Code directory. In Clean there are scripts to run the FERET tests. They use the binaries in ./bin and the sourcce in ./Code. ./Clean contains: README do_new_neighbor feature_search* descriptor do_new_neighbor_old head_align* do_cleanup f_coeffs* head_search* do_gallery f_coeffs_old* headsearch.pl* do_neighbor f_detect* do_neighbor_old f_detect_truth* The basic commands to rerun FERET are: do_gallery do_neighbor Sometimes the hardware platform will require a readjustment of the endianness which can be automatically performed by adding the following into your cshrc file: # for Face pipeline ## set Endian-Swap flag for non-MIPS architecture (default) setenv FLOAT_SWAP off if ( $MACHTYPE != ds && $MACHTYPE != alpha ) then setenv FLOAT_SWAP on endif FERET code was compiled to run on HP platforms. Matlab code was also intended for HPs but was also extensively used on Alpha machines.