
STORY SEGMENTATION (SS) SOFTWARE
--------------------------------

1) Directory contents and description

   Software directories:
   - Source      : The source code of the SS program.
   - Class       : Compiled Class.
   - Lib         : Useful libraries to compile and run.

   Data directories:
   - Data        : The following files must be placed in this directory:
                   * alignment.txt.final - containing the offsets for every video to be processed (see appendix A).
                   * lengthTableTestCollection.txt - containing the video durations (see appendix B)
                   Be sure that every entry in the alignment table has its corresponding entry in the length table.
   - Reference   : It must contain the truth story data (.src_sgm files) of the videos to be processed.
                   Be sure that only the files with an entry in the alignment table are in this directory.
   - Submissions : One directory containing the submission file(s) has to be created for each submission wanted 
                   to be evaluated. Place the storySegmentationSubmission.dtd file into if required.
   - Result      : It contains the result of the submission evaluations. As a result of evaluating a directory 
                   'nameDir' in the Submissions directory, a file named 'ResultnameDir.xml' is created here.

2) How to evaluate a submission?

   Fill the data directories as required.

   Run the script evaluate.sh using the following syntax:

      ./evaluate.sh eps nameDir

      where eps (in seconds) is used to fuzzy match submission times against truth times (T) using a interval [T-eps,T+eps].

   A file named 'ResultnameDir.xml' should appear in the Result directory with the results of the evaluation.

   A video in the submission will not be processed if there is not its corresponding offset in the alignment table.


3) How to compile ?

   The script compile.sh compiles the code from the Source directory using the libraries from the Lib directory and 
   it will place the class files in the Class directory.

4) Getting a result table?

   The Perl script extractRunStory.pl process all the xml files in the Result directory, and output a summarizing table.


Appendix A - Format of the alignment table
-------------------------------------------
One line per video containing the following fields separated by, at least, one blank:

video file name: mpg extension.
offset: in seconds. Precision centiseconds.


Appendix B - Format of the length table
---------------------------------------
One line per video containing the following fields separated by, at least, one blank:

video file name: mpg extension.
video duration time: using the format mm:ss:ccc
