Automatic data capture from scans for amc multiple choice exams.
auto-multiple-choice analyse --projet project-dir [--seuil-coche threshold] [--tol-marque tol] [--list-fichiers files-list.txt | scan-files...]
The AMC-analyse.pl command performs automatic data capture from scans of completed answer sheets of an AMC multiple choice questionnaire. Before calling AMC-analyse.pl, AMC-prepare(1) should have been called to make working documents (--mode s) and the layout should have been processed with AMC-meptex(1).
gives the directory where data files are (see for example AMC-meptex(1)). Defaults to project-dir/data.
gives a directory where to create image reports of the data capture (zooms on the boxes in a zooms sub-directory, name filed extraction as a name-*.jpg file, and layout report as a page-*.jpg file).
When using this option, if --cr is not used, project-cr-dir will be set to project-dir/cr.
Giving the names of the scan files to process can be done in two ways:
these names can be given as arguments,
these names can be written to a plain file (one filename per line), using the --list-fichiers files-list.txt to give the name of this file.
Some parameters for data capture may be set using the following options:
sets the proportion of each box that will be measured to decide if the box is ticked or not. Default value is 0.8.
This parameter is used when converting the grayscale scan file to black and white. To get more black pixels, use a greater value. To get more white pixels, use a smaller value. The threshold must be between 0 and 1. The default value is 0.6.
With this option, only red channel will be used from color scans. This way, all that is written in red will be ignored in the scan. This can be useful when the boxes are printed in red.
defines the tolerance when detecting the four marks in the scans corners. These marks are detected looking for black connected components which dimensions are closed to the target dimension target (exact dimensions of the marks if printing/scanning process where perfectly acurate). If tol is a real number, accepted dimensions are those between (1-tol)*target and (1+tol)*target. If tol is "tinf,tsup" (where tinf and tsup are real numbers), accepted dimensions are those between (1-tinf)*target and (1+tsup)*target. A standard value is 0.2.
This option is to be used when the scans are photocopies from some subjects (different students can have the same subject). In thie case, copy numbers are allocated to students so that their answer sheets, with the same subject number, can be differenciated.
This option can be used in conjonction with --multiple. The copy numbers used for the scans will start from copy_id, in the same order than the scans given as arguments.
gives a file to fill with debugging information.
Alexis Bienven\(:ue <[email protected]>
Translation from French
Translation from French
Copyright © 2008-2013 Alexis Bienven\(:ue
This document can be used according to the terms of the GNU General Public License, version 2 or later.