Arabic
Handwriting Recognition Competition
|
|
|
|
Recognition
of Arabic handwritten words on the basis of the IFN/ENIT-database
The
ICDAR 2007 Arabic Handwriting Recognition Competition aims to bring together
researchers working on Arabic handwriting recognition.
Since
2002 the freely available IFN/ENIT-database is used by more than
30 groups all over the world to develop Arabic handwriting recognition systems.
This competition is the second in a series of competitions to establish the
state of the art of recognizing Arabic handwritten words. Upon making the first
competition at ICDAR 2005 this second competition at ICDAR 2007 gives the
opportunity to further develop methods and discuss results.
A
comparison and discussion of different algorithms and realizations should give
a push in the field of Arabic handwritten word recognition.
Evaluation Process
The object is
to run each Arabic handwritten word recognizer (trained on a part of version
2.0 of the IFN/ENIT-database,
available
(*) Details are
given in the database description (www.ifnenit.com).
A recognizer
may return up to 10 candidates for each classification that not only the first ranked
result can be used for comparison but also the correct result between the 5 or
10 candidates will be used for comparison.
In a first
comparison we use no reject. In the case that a reject is implemented in the
recognizer, the recognition rate in respect to the reject rate will be compared
in a second run.
Running a Recognizer
We run your
recognizer (called myrec) by
invoking it from the command line as follows:
Your recognizer
will be invoked as follows:
myrec
dataset.txt output.txt
The plain
text file formats are as follows:
dataset.txt
The
dataset is now just a list of relative paths to each binary *.tif or *.bmp image to be recognized. For example:
word/1/1.tif
word/1/2.tif
...
output.txt
The output
file should have one line for each input image. Each line should show the name
of the image file that was recognized, followed by the responses (ZIP-Code of
the Tunisian town/village name) for that image.
Each
response is given as a pair of values: the text, followed by the confidence. In
the following example the first line shows that for image word/1/1.tif the
recognizer has produced three word hypotheses: town/village with ZIP-Code 1000,
2000 and 3000, with confidences of 1.0, 0.8 and 0.4 respectively.
word/1/1.tif 1000 1.0
2000 0.8 3000 0.4
word/1/2.tif
Note that the recognizer failed to produce an output
for word/1/2.tif - so we get the file name, but nothing else on that line.
Uploading your Recognizer and Registration
To upload
your recognizer package all the files needed to run it in a zip file.
Then
register and upload the zip file here.
Note
that you may register for and enter the competition even if you do not plan to
attend ICDAR 2007.
Once
you've registered, we can then inform you of any updates.
Implementation
must be Windows or Linux (static linked libs)
executables, or Java .jar archives.
Important Dates
·
Submission of system:
·
Results Announced at ICDAR 2007.
Any Problems
Please contact Volker
Märgner if you have any problems concerning the procedure.