This Matlab software implements the two-dimensional
Scattering Transform and the scattering PCA classifier
described in the paper "Classification with Scattering
Operators".
The main wrapper function is 'generic_classifier',
which comes with detailed help.
The software does not require any installation,
just include the folders (with its subfolders) in the matlab path.
However, the demonstration script requires at least one of the
typical datasets used in classification algorithms.
For that purpose, several retrieval scripts are provided.
They assume a particular arborescence within
the matlab home directory. If you want to use them,
you just need to declare the global variable
'mpath' (for instance in your startup.m) to be
the root directory of the several datasets.
The following datasets are supported.
Unless it is specified otherwise, the datasets are assumed to be
in their original format delivered by their
respective authors. The links to their original
locations are also provided.
-mnist: assumed to be in mpath/mnist
The dataset can be downloaded from http://yann.lecun.com/exdb/mnist/
-usps : assumed to be in mpath/usps.
The dataset in matlab format can be downloaded at
http://www.cmap.polytechnique.fr/scattering/code/usps/usps_original.mat
-caltech: assumed to be in mpath/101_ObjectCategories
The dataset can be downloaded from http://www.vision.caltech.edu/Image_Datasets/Caltech101/
-brodatz: assumed to be in mpath/brodatz_full
The dataset can be downloaded from http://www.ux.uis.no/~tranden/brodatz.html
-CUREt: assumed to be in mpath/curetgrey
The dataset can be downloaded from http://www.robots.ox.ac.uk/~vgg/research/texclass/index.html
-Yaleb (face database). assumed to be in mpath/CroppedYale
The dataset can be downloaded from http://cvc.yale.edu/projects/yalefacesB/yalefacesB.html
Should you have any questions/found any bugs, please
contact me at bruna@cmap.polytechnique.fr
Joan Bruna
December 2010