...............................................................................
International Workshop on Graphics Recognition
Dashed-Line Detection Contest
The Pennsylvania State University
University Park, PA 16802, USA
February 3rd, 1995
...............................................................................
----------------------
Details of the Contest
----------------------
The test images and source code are available on the anonymous ftp site
"ftp.cse.psu.edu" (130.203.2.11).
The path is : pub/vision/dashed_lines/(src, images)
The images directory contains samples of simple, medium and complex
dashed-line images in TIFF 5.0 format. The khoros routines "raw2viff"
and "viff2tiff" have been used to convert raw images into TIFF images.
The src directory contains all the source files for the creation of
these images. Any updates on the src files will be posted.
The README file in the ftp site directory contains information about the
dashed-line parameters (segment length, gap length etc.), image sizes,
complexities, etc.
Contestants will have to ftp their software onto Sun SPARC class of machines
which will be provided. They will have to run a live demonstration of
their program on the TIFF images which will be generated randomly at
the time of the contest. Those who do not have access to the ftp site
should contact the concerned persons at email addresses given below
for information on alternate ways of obtaining the images or src code.
Prizes will be awarded to winning entries at the workshop.
----------
Evaluation
----------
Programs will be judged on the basis of the following performance metrics :
Ability to
- detect all the dashed-lines.
- accurately determine the pixel locations of the dashed-lines
like start point, end point.
- track each dashed-line correctly.
- determine accurately the segment lengths, gap lengths and thickness.
- determine the different kinds of dashed-lines in an image (in case of
the medium and complex images).
- identify the objects formed by dashed-lines (dashed rectangular boxes,
dashed triangular shapes, dashed-hatching lines, etc.) (again in
case of the medium and complex images).
- segment the text strings and continuous lines from the
dashed-lines (in case of complex images).
- track the dashed curves accurately (again in case of complex images).
NOTE : The exact evaluation methods are not yet complete and will be
posted as soon as they are ready.
For further correspondence, send a mail to Rangachar Kasturi (alternative mailer if your browser understands mailto) or to Arathi Prasad (alternative mailer if your browser understands mailto).