68 lines
2.6 KiB
Text
68 lines
2.6 KiB
Text
|
Note: this info is related to example files, used to test gOCR. As of this
|
||
|
|
writing, these files are not available to non-developers. So, if you aren't
|
||
|
|
a developer, forget about this file.
|
||
|
|
|
||
|
|
EXAMPLE FILES
|
||
|
|
|
||
|
|
1. Scanning
|
||
|
|
The examples can be scanned from anything; when looking for something, try to
|
||
|
|
have in mind the kind of tests you are expecting to do: if you're testing
|
||
|
|
accents recognition, look for texts in portuguese, french, etc. (pretty obvious,
|
||
|
|
but keeping this in mind will help to have a large gamma of files covering
|
||
|
|
different kinds of tests).
|
||
|
|
|
||
|
|
If you're not interested in testing DPIs, scan at 150 or 300dpi.
|
||
|
|
|
||
|
|
If you're not interested in testing the dust removal, cleaning, etc, functions,
|
||
|
|
do the best scan you can. Usually increasing brightness and contrast will
|
||
|
|
provide a sharper, cleaner image.
|
||
|
|
|
||
|
|
Save the image in a supported format: for example, pgm or jpg. If a compression
|
||
|
|
will result in a significant reduction of size, compress the image. BZIP2
|
||
|
|
usually is the best compressor around, but gzip is more popular in the unix
|
||
|
|
world. In the wintel world, people use ZIP, and usually will have to search for
|
||
|
|
an application capable of opening .gz or .bz2 (though WinZIP opens at least the
|
||
|
|
former).
|
||
|
|
|
||
|
|
2. Sorting
|
||
|
|
To help others to find the files they are looking for, the examples/ directory
|
||
|
|
is divided in several other directories, which may be subdivided. When
|
||
|
|
uploading a new example, look for the most suitable location. Depending of the
|
||
|
|
directory, you probably will name your file with interesting info: for example,
|
||
|
|
when uploading a image with all the characters of the foo font, the best thing
|
||
|
|
to do is to place it at examples/fonts/foo.jpg.
|
||
|
|
|
||
|
|
3. "Translation"
|
||
|
|
Along with the image file, upload a text file with the expected output. Be
|
||
|
|
careful with this file: it must resemble the original text as much as possible.
|
||
|
|
Don't add extra new lines (\n), keep hyphenized words, etc. Name this file with
|
||
|
|
the same name of the image file.
|
||
|
|
|
||
|
|
In the beginning of the text file, you should provide comments, to help
|
||
|
|
searches. Use the following sample:
|
||
|
|
|
||
|
|
# Comments
|
||
|
|
# DPI:
|
||
|
|
# Colors:
|
||
|
|
# Image size (colsXrows):
|
||
|
|
# Fonts:
|
||
|
|
# Font sizes:
|
||
|
|
# Layout form:
|
||
|
|
# Number of pictures:
|
||
|
|
# Language:
|
||
|
|
# Quality of scan:
|
||
|
|
# Non-ASCII characters:
|
||
|
|
# Extra:
|
||
|
|
|
||
|
|
Check existing examples to see what people have been doing.
|
||
|
|
|
||
|
|
Any lines that begin with # will be considered comments, so you may use several
|
||
|
|
lines for comments or add new fields. Though gOCR itself doesn't depend on, and
|
||
|
|
won't use, this file, it will be used by scripts.
|
||
|
|
|
||
|
|
4. Other sources (WEB)
|
||
|
|
|
||
|
|
- http://www.clerkweb.house.gov/elections/elections.htm (Nov2002)
|
||
|
|
PDF-files with lot of tables
|
||
|
|
|