This is an implementation of the 'bypass' version of the standard model. 
The implementation follows "Serre, T. and Wolf, L. and Poggio, T.", 
"A new biologically motivated framework for robust object recognition",CVPR,2005.  
There are other implementations (C,C++,matlab) available from http://cbcl.mit.edu/cbcl . 
This version allows you to arbitrarily replace each layer with several faster 
approximations. For questions and bug-reports please contact  sharat@mit.edu

The following lists the important files in the implementation. 

BASELINE-CORE
--------
read_all_images    - loads and normalizes a whole directory of images. Useful to training/testing
default_options.m  - sets the default filter parameters. 
init_gabor.m       - initializes the bank of gabor filters
s1_baseline.m      - generates the output of the S1 units using the filterbank and options specified
c1_baseline.m      - generates the C1 unit activations by max-pooling the S1 units. The spatial and 
                     scales covered by each unit is specified by the options parameter
get_random_patches - (used only during training) generates the dictionary of features used for 
                     recognition. 
c2_baseline.m      - generates the C2b unit activation. There are as many units in this layer as
                     the number of prototype patches used. 

BASELINE-HELPER
----------------
patch_response.m   - compute the response of C1 activation to a single prototype feature
normalize_image.m  - normalizes range and size of the image. 
time_c1_c2.m       - times the response of C1 and C2 computation
visualize_hmax_features

BASELINE-WRAPPERS
-----------------
callback_c2_baseline.m - a wrapper that uses default options to generate C2b unit activations. 
demo_baseline.m        - a demo script showing how to use the model for classification.
c2_basleine_for_set    - computes c2 features using baseline implementation for a whole set of images


S1 APPROX-SEPARABLE
-------------------
separable_options.m    - use this instead of default_options
init_separable.m       - use this instead of init_gabor
s1_separable.m         - use this instead of s1_baseline

S1 APPROX-SVD
------------------
svd_options.m          - just calls default_options
init_svd.m             - initialize SVD approximations to the filters. Use this instead of init_gabor.m
s1_svd                 - use this instead of s1_baseline

S1 APPROX-BOX
------------------
box_options           -  use this instead of default_options
init_box              -  use this instead of init_gabor
s1_box                -  use this instead of s1_baseline
integral_image	      -  helper function
rot_integral_image    -  helper function

C2 APPROX - subsample
--------------------
callback_c2_subsampling- use this instead of callback_c2_baseline
c2_subsample           - use this instead of c2_baseline in stand alone code
subsample_patch_response-helper function
C2 APPROX-PCA
--------------------
callback_c2_pca        - use this instead of callback_c2_baseline
c2_pca                 - use this instead of c2_baseline in stand alone code
patches_baseline_pca_01- pca decomposition of the patches_baseline_01