The code is structured as follows:
bow - code for bag-of-words based classifier
egtd - code for computing EGTD descriptor and classifier
stacking - code for fusion of descriptors
work - working folder for data files

Folder bow contains code for bag-of-words model. It is adapted from
VLFeat demo by Andrea Vedaldi and Brian Fulkerson.
Script do_all.m calls the three modules:
- makeCodebook.m is used for codebook creation
- extract_features.m computes bag-of-words representation for all images
- classify_bow_auto.m trains and tests SVM classifier for multiple random 
  dataset splits

In each module there are a few parameters that need to be set. The most
important are the paths to the images and working folder: conf.imageDirRGB and
conf.workDir. In conf.phowOpts are settings for SIFT descriptor computation.
Please refer to VLFeat documentation for further information.

Folder egtd contains code for computing EGTD descriptor.
Script do_all.m call the two modules:
- ucmerced_index.m computes EGTD for images from the UC Merced dataset
- ucmerced_classify_svm_auto.m trains and tests SVM classifier for multiple
  random dataset splits

Paths to the images and working folder are in variables PREFIX and DESCDIR,
respectively. It is also possible to set the parameters of the Gabor filter
bank.

Folder stacking contains code for fusion of descriptors.
Script stacking_bow_svm_nfold_21class.m trains and tests the classifier.

Paths to the images and working folder are in variables PREFIX and DESCDIR,
respectively. Ntrain_class is the number of training images from each class,
nfolds is the number of folds of the training set for computing the mid-level
representation. Nruns is the number of different random training/test splits,
and beta is the parameter of the softmax function.

The code uses VLFeat library (vlfeat.org), as well as LIBSVM (http://www.csie.ntu.edu.tw/~cjlin/libsvm/) and liblinear (http://www.csie.ntu.edu.tw/~cjlin/liblinear/) libraries.

-----------------------------------------------------------------------
"This software is made available under the Creative Commons 
Attribution-Noncommercial License.  You are free to use, copy, modify, and 
re-distribute the work.  However, you must attribute any re-distribution or 
adaptation in the manner specified below, and you may not use this work for 
commercial purposes without the permission of the author.

Any re-distribution or adaptation of this work must contain the author's name 
and a link to the software's original webpage.  For example, any 
re-distribution of the software must contain a link to:
http://dsp.etfbl.net/aerial/

This software comes with no guarantees, and all use of these codes is
entirely at the user's own risk."


