- Meeting MBARI 25 August 2004 - 
## classification of particle categories ##


Content of folder CODE_CLASSIFIER:
. folder DB -> toy database to run the program
. folder netlab -> files belonging to NetLab package that are used (i.e. to estimate MoG)
. run_manyExp.m -> script file to run experiments of classification using three classes (see DB folder)
. training_ljmNL3.m -> function to get training models
. test_ljmNL3.m -> function to test data
. collect_data_classes.m -> script to compute image features (it could/should be done during training and test)
. calcola_invarianti.m -> function that computes Schmid invariants
. apply_non_lin3.m -> function that apply a non-linearity to the invariants
. comp_stat.m -> function that estimates the mean of the background
. fisher_l_d.m -> function that implements Fisher Linear Discriminants
. ERvsParam.m -> example of script to tune the system in order to find the optimal nr. of dimension of feature space, nr. of components in the mixture and kind of their covariance matrix
. find_initcond.m -> example of script to find a good intialization for the random number generator (used in the MoG estimate)


#$%#$%#$%#$%#$%#$%#$%#$%#$%#$%#$%#$%#$%#$%#$%#$%

EXAMPLE:
>> collect_data_classes
cl 1|
cl 2|
cl 3|
>> run_manyExp
@@@@@@@@@@ Experiment nr.1 @@@@@@@@@@ 
TRAINING 
Computing FLD 
Computing MoG 
computing MoG-1 
computing MoG-2 
computing MoG-3 
TEST 
cl BACT:/ 93 on 100 
cl RBC:/ 100 on 100 
cl SPRM:/ 100 on 100 
TEST 
cl BACT:/ 94 on 100 
cl RBC:/ 100 on 100 
cl SPRM:/ 98 on 100 

#$%#$%#$%#$%#$%#$%#$%#$%#$%#$%#$%#$%#$%#$%#$%#$%


- Calteth - 

Marc'Aurelio Ranzato 
aurelio@vision.caltech.edu