DEPENDENCES
===========
LIBOCAS requires only standard C libraries. 


INSTALL
=======

Unpack the library to a folder of your choice, jump to the folder and issue 

    make

which should produce

    svmocas         ... standalone application for training binary linear SVM classifiers
    msvmocas        ... standalone application for training multi-class linear SVM classifiers
    linclassif      ... implementation of linear classification rule
    svmocas.so      ... Linux library

In addition, if Matlab mex compiler is in path then the following MEX functions will be generated
    
    msvmocas.mexXXX          ... Training multi-class linear SVM classifier
    msvmocas_light.mexXXX    ... Training multi-class linear SVM classifier from SVM^light file
    svmocas.mexXXX           ... Training two-class linear SVM classifier
    svmocas_lbp.mexXXX       ... Training two-class linear SVM classifier for grey-scale images
    svmocas_light.mexXXX     ... Training two-class linear SVM classifier from SVM^light file
    linclassif_light.mexXXX  ... Linear classifier loading examples directly form SVM^light file

    compute_auc              ... Computes area under ROC.
    lbppyr_features.mexXXX   ... Computing LBP feature descriptor for given images.


MATLAB
======

First, CD to the root folder of LIBOCAS and then:

To get list of all implemented functions type 
  help Contents

Each of the implemented SVM solvers has its detailed help together with a simple example of use, 
just try
  help svmocas
  help svmocas_light
  help svmocas_lbp
  help msvmocas
  help msvmocas_light

To test all implemented SVM solvers type
    libocas_test

To test SVMOCAS_LBP for training translation invariant image classifiers try
    svmocas_lbp_example

To get help type 
    help Content
    

STANDALONE APPLICATIONS
=======================

To get help type
    ./svmocas
    ./msvmocas
    ./linclass


TROUBLESHOOTING
===============

Do not hasitate to send us email
  xfrancv@cmp.felk.cvut.cz