CC

Open source Image category classifier

Usage:

1. create the vocabulary and save it in a file.

      ./run.py -v <images_folder> -o <vocab_output_file>
      
    <images_folder>: is a path to a folder containing the images that will construct
                     the vocabulary from them.
    <vocab_output_file>: is a path to a file where the vocabulary will be saved in.
   

2. train the classifier and save it for later use.

      ./run.py -t <train_folder> -r <ref_vocab_file> -o <classifier_output_file> -d <categories_dictionary_output_file>
    
    <train_folder>: is a path to a folder where the training images will be found.
                    it should have a sub folder for each category named after its label.
                    Note: if "Cow" and "cow" were 2 labels they will be considered the same.
    <ref_vocab_file>: is a path to a file where the vocabulary saved in step 1.
    <classifier_output_file>: is a path to a file where the trained classifier will be saved in.  
    <categories_dictionary_output_file>: is a path to a file where the categories dictionary will be saved in.
   

3. test and evaluate the performance of the classifier.

      ./run.py <evaluation_type> <test_folder> -r <ref_vocab_file> -c <ref_classifier_file> -d <ref_categories_dictionary_file>
      
    <evaluation_type>: use -e for printing counts for wrong predicitons.
                       use -s for printing precision scores and their mean.
    <test_folder>: is a path to a folder where the testing images will be found.
                   it should have a sub folder for each category named after its label.
                   this label will be used to determine the correctness of the classifier's prediction.
    <ref_vocab_file>: is a path to a file where the vocabulary saved in step 1.
    <classifier_output_file>: is a path to a file where the trained classifier saved in step 2.  
    <categories_dictionary_output_file>: is a path to a file where the categories dictionary saved in step 2.