Bayes classifier

Binary classification of images using Bayes classifier

Usage

There are defined 3 usage cases

1. Evaluate trained classifier on the test set

• ./bayes --evaluate --test p1.txt n1.txt --train p2.txt n2.txt --threshold NUM [...]

2. Get table which contains precision and recall for possible threshold values (computed using cross-validation)

• ./bayes --analyze --train pos.txt neg.txt [--q 2^NUM] [--method BAYESIAN_RGB | --method BAYESIAN_R] [--subsample]

3. Calculate a probability for image img.bmp (only .bmp format supported)

• ./bayes --predict --train pos.txt neg.txt --image img.bmp [--q 2^NUM] [--method BAYESIAN_RGB | --method BAYESIAN_R] [--subsample]

Command line arguments

Run ./bayes VARIANT INPUT OPTIONAL where

• VARIANT

• --evaluate: evaluation of implemented method

• --analyze: show table of rates for training samples

• --predict: predict probability for sample using defined threshold

• INPUT

• --test positive.txt negative.txt

• --train positive.txt negative.txt

• OPTIONAL

• --method: possible values BAYESIAN_R or BAYESIAN_RGB (default is BAYESIAN_RGB)

• --q NUM: change size of histogram dimensions (default 16)

• --subsample: subsample images to descrease exec time (default not use)

Examples

• ./bayes --evaluate --threshold 0.37 --subsample
• ./bayes --evaluate --train p1.txt n1.txt --test p2.txt n2.txt --threshold 0.34
• ./bayes --analyze --train p.txt n.txt
• ./bayes --train p1.txt n1.txt --test --image img.bmp