Eyegrade User Manual

Author: Jesús Arias Fisteus

Contents

This user manual refers to Eyegrade 0.2 and later versions. For the 0.1.x series see this other user manual.

1   Installing Eyegrade

Eyegrade depends on the following free-software projects:

1.1   Upgrading from Eyegrade 0.1.x to Eyegrade 0.2

If you have any version of the Eyegrade 0.1.x series already installed in your system, you have to install PyQt.

In Linux, install the package python-qt4 and update Eyegrade following the instructions at Updating Eyegrade.

In Windows, Download PyQt and install it. If you installed Eyegrade 0.1.x following the instructions in this user manual, you have Python 2.6-x86. Choose that version of the installer of PyQt. Alternatively, there is a copy of the file you need at PyQt-Py2.6-x86-gpl-4.9.6-1.exe. The installer of PyQt will probably detect you Python installation automatically.

After that, update Eyegrade following the instructions at Updating Eyegrade.

The directory layout of the code of Eyegrade has changed with respect to versions 0.1.x. The code is now inside the eyegrade subdirectory instead of the src/eyegrade subdirectory. Therefore, the PYTHONPATH environment variable should be set to the main directory of Eyegrade now, instead of the src subdirectory as before. When upgrading, remember to update this environment variable if you have configured it in your system-wide environment variables.

If you cannot see the icons of the buttons the first time you run the new version of Eyegrade, reboot the computer and try again.

1.1.1   Main changes from versions 0.1.x

The main changes from Eyegrade 0.1.x to Eyegrade 0.2 you need to be aware of are:

  • The main program is run now without command line arguments. You'll select the .eye file, student lists, webcam to use, etc. from the user interface.
  • There is the new concept of a session. A session stores the state of the grading of a given exam. You create a new session when you start to grade the exam. The images of the exams and results are stored in this session as you grade them. You can later resume a session and continue grading it. A session is stored in the directory of your file system that you choose.

1.2   Installation on GNU/Linux

If your Linux distribution is not very old, it should provide most of the needed software packages. Specific instructions for Debian GNU/Linux and Ubuntu are provided below.

1.2.1   Installation on Debian and Ubuntu

Almost all the required software packages are already available in recent versions of Debian GNU/Linux and Ubuntu. The only exception are the Python bindings for Tre, which have to be installed manually.

Using your favorite package manager (apt-get, aptitude, synaptic, etc.), install the following packages: python (check that the version is either 2.6 or 2.7), python-qt4, python-opencv, python-numpy, libavformat53, libtre5. In older versions of Ubuntu and Debian, you might need to install also libcv2.1 (in even older versions, the name of this package is libcv4 instead). If you can't find libavformat53 in your distribution, use libavformat52 instead.

Then, you have to install the Python bindings for Tre. First, install these two additional packages: python-dev, libtre-dev. Then, from the command line, download, compile and install the Python bindings:

wget http://laurikari.net/tre/tre-0.8.0.tar.gz
tar xzvf tre-0.8.0.tar.gz
cd tre-0.8.0/python/
python setup.py build
sudo python setup.py install

Now, you only need to download Eyegrade using the git source code revision system (install the git package if you do not have it):

cd $DIR
git clone -b master git://github.com/jfisteus/eyegrade.git

Note: replace $DIR above with the directory in which you want Eyegrade to be installed.

Finally, add the $DIR/eyegrade directory to your PYTHONPATH and check that Eyegrade works:

export PYTHONPATH=$DIR/eyegrade
python -m eyegrade.eyegrade

The export command works only in the current terminal. You can make it permanent by adding it to your $HOME/.bashrc file (if you use the BASH shell).

That's all! Eyegrade should now be installed. For further testing, go to Launching Eyegrade.

1.3   Installation on Microsoft Windows

You have to follow these steps, explained in the following sections, in order to install Eyegrade in Windows:

1.- Install Python 2.6 (including Tre).

2.- Install PyQt.

3.- Install OpenCV 2.1.

4.- Install Eyegrade itself.

1.3.1   Installing Python

The easiest way to install Python, PyQt and Tre in Windows is to download a ZIP file that contains all of them and extract it in your file system.

1.- Download the ZIP file from: Python26.zip.

2.- Extract it somewhere in your file system (I recommend C:\). A directory named Python26 will appear. Be aware that the full path of the directory where you extract it cannot contain white-spaces.

3.- Add the main directory (Python26) of your Python installation to your system PATH. For example, if you uncompressed Python at C:\, add C:\Python26 to the system PATH variable.

You can test your installation by opening a new command line console and launching the interactive Python interpreter in it:

Python

If it does not start, you have probably not added it correctly to your system PATH. Opening a new console is important because changes in the system PATH apply only to newly-opened consoles.

Once in the Python interpreter, the following command should work:

import tre

This command should not output any message. If it does, there is a problem with the installation. If tre complains about a missing DLL, the problem is probably that the installation directory of Python is not in the system PATH.

If you already have a Python 2.6 installation and want to use it, you must, on that installation of Python, download and install Tre 0.8.0. You will need Microsoft Visual Studio 2008 (the express version is free and works) for this last step.

1.3.2   Installing PyQt4

Download PyQt. Select the Windows 32-bit installer for Python 2.6, event if you have a 64-bit version of Windows. Alternatively, there is a copy of the file you need at PyQt-Py2.6-x86-gpl-4.9.6-1.exe.

Run the installer. From the optional software that the installer suggests, you only need to select the Qt runtime.

1.3.3   Installing OpenCV

Download the EXE installer of OpenCV 2.1.0 for Windows platforms: OpenCV-2.1.0-win32-vs2008.exe. There is a copy of the same file at OpenCV21.exe.

Execute the installer. Again, it is better to choose an installation path which has no white-spaces in it. The installer will eventually ask to put OpenCV in your system PATH. Answer yes for this user or yes for all the users.

In order to test the installation, open a new command prompt window (it must necessarily be a new window for the system path to be updated). Run the python interpreter as explained in the previous section and type in it:

import cv

This command should not output any message. If it does, there is a problem with the installation.

1.3.4   Installing Eyegrade

By now, the recommended way to install Eyegrade is through the Git version control system. This way it will be easier to update Eyegrade in the future, when new versions are released (see Updating Eyegrade).

In order to install Eyegrade through Git, follow these steps:

1.- Download and install Git if you do not have it installed. The installer and installation instructions are available at <http://git-scm.com/>.

2.- Open a command line prompt (for example, a Git shell), enter the directory you want Eyegrade to be installed (again, with no white-spaces in it), and type:

git clone -b master git://github.com/jfisteus/eyegrade.git

If you prefer not to install Git:

1.- Download the ZIP file eyegrade-0.2.6.zip. Extract it in your file system, in a directory with no white-spaces in its path.

Once you have Eyegrade installed (either with or without Git), test it. For example, if you have installed both Python and Eyegrade at C:\:

set PYTHONPATH=C:\eyegrade
C:\Python26\python -m eyegrade.eyegrade

It should dump a help message.

Tip: it may be convenient adding C:Python26 to your system path permanently, and adding PYTHONPATH to the system-wide environment variables. There are plenty of resources in the Web that explain how to do this. For example, http://www.windows7hacker.com/index.php/2010/05/how-to-addedit-environment-variables-in-windows-7/.

Eyegrade should now be installed. Nevertheless, it might be a good idea to reboot now the computer, in order to guarantee that the installation of OpenCV and PyQt has completed. After that, go to Launching Eyegrade.

1.4   Installation on Mac OS X

Sorry, Eyegrade is not currently supported on that platform. Volunteers to support the platform are welcome.

1.5   Updating Eyegrade

From time to time, a new release of Eyegrade may appear. If you installed Eyegrade using Git, updating is simple. Open a command prompt window, enter the Eyegrade installation directory and type:

git pull

This should work on any platform (Linux, Windows, etc.)

2   Grading Exams

The main purpose of Eyegrade is grading exams. In order to grade exams, you will need:

2.1   Launching Eyegrade

This section explains how to run Eyegrade. If it is the first time you use Eyegrade, you can try it with the sample file exam-A.pdf located inside the directory doc/sample-files of your installation of Eyegrade. Print it. You'll find also in that directory the file exam.eye that contains the metadata for this exam. You'll need to load this file later from Eyegrade.

Eyegrade can be launched from command line:

python -m eyegrade.eyegrade

This command opens the user interface of Eyegrade:

Eyegrade main window

Before beginning to grade exams, especially the first time you run Eyegrade, you can check that Eyegrade can access your webcam. In the Tools menu select the Select camera entry:

Select camera dialog

The next step is creating a grading session. Select New session in the menu Session. A multi-step dialog will ask for some data Eyegrade needs for creating the session:

  • Directory and exam configuration: you need to enter here the following information:
    • Directory: select or create a directory for this session. The directory must be empty.
    • Exam configuration file: select the .eye file associated to this exam. If you printed the sample exam distributed with Eyegrade, use the exam.eye file from the same directory.
  • Student id files: select zero, one or more files that contain the list of students in the class. The files should be plain text and contain a line per student. Each line must have a first field with the student id and, optionally, a second field with the student name. It may have more fields, which Eyegrade will ignore. Fields must be separated by one tabulator character.
  • Scores for correct and incorrect answers: this step is optional. If you provide the scores awarded to correct answers (and optionally deducted from incorrect answers), Eyegrade will show the marks of each exam.

After you finish with this dialog, Eyegrade opens the session. It shows the image from the webcam and starts scanning for the exam. Point the camera to the exam until the image is locked. At this point, Eyegrade should show the answers it has detected. Read the following sections for further instructions.

2.2   The session directory

A grading session in Eyegrade represents the grading of a specific exam for a group of students. For example, you would grade the exams for the final exam of all your students in the subject Computer Networks in just one session. Other exams, such as the re-sit exam of the same subject, should go in separate sessions.

Grading sessions are associated to a directory in your computer. You select or create this directory when you create a new session. Eyegrade stores there all the data belonging to the grading session (configuration file, student lists, grades, images of the already graded exams, etc.)

You can open again later an existing session with the Open session option of the Session menu. In the file selection dialog that appears, select the session.eye file inside the directory of the session you want to open. When you open the session, you can continue grading new exams that belong to that session.

2.3   Application modes

At a given instant, the application is in one of these modes:

  • No session mode: no session is opened. You can open an existing session or create a new session.
  • Search mode: a session is open. The application continually scans the input from the webcam, looking for a correct detection of an exam.
  • Review mode: a session is open. The application shows a still capture of an exam with the result of the grading, so that the user can review it and fix things, if necessary, before saving the score of the exam.
  • Manual detection mode: a session is open. In the rare cases in which the system is not able to detect the geometry of the exam, you can enter this mode and mark the corners of the answer tables. Eyegrade will be able to detect the tables once you tell it where the corners are.

The application starts with no open session. Once you open or create a session, it changes to the search mode. When the system detects an answer sheet that it can read, it locks the capture and enters the review mode. Once you save the score of the exam, Eyegrade automatically goes back to the search mode in order to scan the next exam.

You can enter the manual detection mode by issuing the appropriate command while in the other modes.

From any of the other modes, you can go back to the no session mode with the Close session command in the Session menu.

2.4   The search mode

In the search mode, you have to get the camera to point to the answer table of the exam, including, if present, the id box above it and the small squares at the bottom.

Eyegrade will continually scan the input of the webcam until the whole exam is correctly detected. At that moment, Eyegrade will switch to the review mode.

Sometimes, Eyegrade is able to detect the answer table but not the ID table at the top of it. You can notice that because the detected answers are temporary shown on top of the image. At this point, you may try further until the ID box is also detected, or just use the Capture the current image command of the Grading menu, which will force the system to switch to the review mode, using the most recent capture in which the answer table was detected. You will be able to manually enter the missing student id in that mode.

In rare occasions, Eyegrade could fail event to detect the answer table. The Manual detection command of the Grading menu allows you to help the system detect it.

These are the commands available in the search mode, all of them at the Grading menu:

  • Capture the current image (shortcut 's'): forces the system to enter the review mode with the the most recent capture in which Eyegrade was able to detect the answer table. If there is no such capture, the system just uses the current capture.
  • Manual detection of answer tables (shortcut 'm'): the system enters the manual detection mode, in which you can help the system detect the answer table by marking the corners of the answer tables. After that, the system will detect the answers of the student and automatically enter the review mode. See The manual detection mode.

2.5   The review mode

In the review mode you can review and, if necessary, fix the information detected by Eyegrade in the current exam. You can do it on both the answers given by the student to each question and the student id. You enter the review mode in one of the following three different situations:

  • With the answers of the student and her id detected. This is the usual case. Eyegrade was able to detect the whole exam, and you can review the information extracted from it.
  • With the answers of the student, but without her id. This is the case when you use the Capture the current image command in the search mode because Eyegrade detected the answer table in at least one capture, but not the student id box. In this case, you can review the answers given by the student and manually enter her id.
  • With neither the answers of the student nor her id. This is the case when you use the Capture the current image command in the search mode because Eyegrade was not able to detect anything from the exam. In this situation, you can switch to the manual detection mode to help the system to detect the answer tables, and manually enter the student id.

The user interface shows, in this mode, a capture of the exam augmented with the detected information, as shown in the following image:

Eyegrade in the review mode

As you can see, the system shows:

  • The answers of the student, with a green circle for correct answers and a red circle for the incorrect ones. When the student leaves a question unanswered, or provides a wrong answer for it, the correct answer for that question is marked with a small dot.
  • The detected student id, at the bottom of the image, and his name (when the name is provided in the student list files).
  • The total number of correct, incorrect and blank answers, at the bottom. The total score of the exam is also shown if the session is configured with the scores for the answers.
  • The model of the exam. The model is detected from the small black squares that are printed below the answer table.
  • The sequence number of this exam. It is incremented with each graded exam.

In this mode, you can perform the following actions (see the Grading menu):

  • Modify the answers of the student, if there are mistakes in the automatically-detected answers, as explained in Modifying student answers.
  • Modify the student id, if the system did not recognize it or recognized a wrong id, as explained in Modifying the student id.
  • Save and capture next exam (shortcut 'Space-bar'): saves the grades of this exam as well as the annotated captured image, and enters the search mode in order to detect the next exam. Tip: before saving, it is better to remove the exam from the sight of the camera to avoid it from being captured again. You can even put the next exam under the camera before saving to speed up the process.
  • Discard capture (shortcut 'Backspace'): discards the current capture without saving it. It is useful, for example, when the capture is not good enough, or when you discover that the same exam has already been graded before.
  • Manual detection of answer tables (shortcut 'm'): the system enters the manual detection mode, in which you can help the system detect the answer table by marking the corners of the answer tables. After that, the system will detect the answers of the student and automatically enter again the review mode. This command is allowed only when the system failed to recognize the geometry of the answer tables. See The manual detection mode.

2.5.1   Modifying student answers

The optical recognition system of Eyegrade may fail sometimes, due to its own limitations, or students filling their exams in messy ways. Sometimes, Eyegrade shows a cell in the answer table as marked when it is not, or a cell is not marked when it actually is. In addition, if Eyegrade thinks that two cells of the same question are marked, it will leave that question as blank.

You are able to fix those mistakes at the review mode. Click on a cell of the answer table to change an answer of the student that was not correctly detected by Eyegrade: when the student marked a given cell, but the system detected the question as blank, or simply showed other answer of that question as marked, just click on the cell the student actually marked. When the student left a question blank but the system did mark one of the cells as the answer, click on that cell to clear it. In both cases, Eyegrade will compute the scores again and immediately update the information on the screen.

2.5.2   Modifying the student id

Normally, you should provide Eyegrade with the list of class, because detection of student ids performs much better in that case. When scanning the id in an exam, Eyegrade sorts ids of the students in class according to the estimated probability of being the id in the exam. The one with the most probability is shown.

In the review mode, you can enter the correct student id when Eyegrade does not detect it, or detects a wrong one. When you select the Edit student id command in the Grading menu, a dialog for selecting the student id is shown:

Dialog for changing the student id

The dialog shows the students from the student list sorted by their probability (according to the OCR module) of being the student whose id is in the exam. You just choose one in the drop-down menu. In addition, you can filter students by writing part of their id number or their name.

If the student is not in your list, you can also enter in the dialog her id number and name. If you do that, follow the same format: student id, white space, student name.

2.6   The manual detection mode

In some rare occasions, Eyegrade may not be able to detect the answer tables. In those cases, you can enter the manual detection mode from the search mode (and also from the review mode if you entered that mode using the Capture the current image command). When entering the manual detection mode, the latest capture of the camera will be shown.

In this mode, just click with the cursor in the four corners of each answer table (a small circle will appear in every location you click). The order in which you click on the corners does not matter. After having done that, Eyegrade will infer the limits of each cell, and based on them it will read the answers of the student and the exam model. It will enter then the review mode.

The following two images show an example. In the first image, the user has selected six corners (notice the small blue circles):

Eyegrade in the review mode

After she selects the remaining two corners, the system detects the answers and goes back to the review mode:

Eyegrade in the review mode

Note, however, that the student id will not be detected when you use this mode. When the system goes back to the review mode, set the id as explained in Modifying the student id.

At any point of the process, you can use the Manual detection of answer tables command (shortcut 'm') to reset the selection of corners and start again. If you think that the captured image is not good enough, you can also use the discard command (shortcut 'Backspace') to go again to the search mode.

Tip: in the manual detection mode, make sure that the captured image shows all the answer tables as well as the exam model squares at the bottom.

3   Processing Student Grades

The output produced by Eyegrade consists of:

3.1   The answers file

The file eyegrade-answers.csv produced by Eyegrade contains the scores in CSV format (with tabulator instead of comma as a separator), so that it can be easily imported from other programs such as spreadsheets. This is an example of such a file:

0   100999991       D       9       6       0       1/2/2/4/1/2/2/0/0/3/2/0/3/2/0/4/3/0/1/2
1   100999997       C       15      1       0       2/4/4/3/1/0/1/2/1/1/0/1/0/4/3/0/1/4/3/4
2   100800003       D       6       14      0       4/2/2/2/1/2/1/3/2/1/3/1/2/1/3/1/4/1/4/3
3   100777777       A       7       13      0       3/2/3/2/3/3/2/4/3/1/3/1/4/1/4/2/2/3/4/2

The columns of this file represent:

1.- The exam sequence number (the same number the user interface shows below the student id in the review mode).

2.- The student id (or '-1' if the student id is unknown).

3.- The exam model ('A', 'B', 'C', etc.)

4.- The number of correct answers.

5.- The number of incorrect answers.

6.- The number of undetermined answers (answers marked as blank because of the system detecting more than one marked cell).

7.- The response of the student to each question in the exam, from the first question in her model to the last. '0' means a blank answer. '1', '2', etc. mean the first choice, second choice, etc., in the order they were presented in her exam model.

Exams are in the same sequence they were graded. See Exporting a listing of scores to know how to produce a listing of scores in the order that best fits your needs.

Tip: if you start a new grading session from the same directory, the file eyegrade-answers.csv will not be overwritten. New grades will just be appended at the end. Thus, it is safe stopping a grading session, closing the application and continuing later. Separate grading sessions must be executed from different directories to avoid using the same eyegrade-answers.csv file.

Tip: you can edit this file with a text editor if, for example, you discover that the same exam was graded more than once (just remove the duplicate line).

3.2   Exporting a listing of scores

You will probably want to import the listing of scores from your grade-book. You can easily process eyegrade-answers.csv to produce a CSV-formatted file with three columns: student id, number of correct answers and number of incorrect answers, in the order you want. You can even produce the listing to for just a subset of the students.

In order to do that, you need a listing of students whose grades you want to list. The listing must be a CSV file in which the first column contains the student ids (the rest of the columns will be just ignored). Normally, you will use the same listing of students you used to run Eyegrade. This is an example of such a file:

100000333    Baggins, Frodo
100777777    Bunny, Bugs
100999997    Bux, Bastian B.
100999991    Potter, Harry
100800003    Simpson, Lisa

This command will produce the listing in a file named sorted-listing.csv:

python -m eyegrade.mix_grades eyegrade-answers.csv student-list.csv -o sorted-listing.csv

The output for the listing above, and the sample file shown in The answers file, would be:

100000333
100777777    7      13
100999997    15     1
100999991    9      6
100800003    7      13

Scores will be in the same order as the student list. The second and third columns represent the number of correct and wrong answers, respectively. In the example, the first student has those columns empty because there is no exam associated to his id.

Importing the previous file in a spreadsheet should be straightforward, because the list of students will now be in the same order as your spreadsheet.

If there are exams in the answers file of students not in your list, the default behavior is including them in the listing, after the rest of the students. The rationale behind this behavior is apreventing accidental losses of student scores. This behavior can be changed (see Exporting a listing for a subset of students).

See Mixing manually-graded questions if you need to produce listings in exams combining MCQ questions with manually-graded questions.

3.2.1   Exporting a listing for a subset of students

In order to extract the scores for just a subset of the students, create a student list with the ids of the students you want and run the program with the -i option:

python -m eyegrade.mix_grades eyegrade-answers.csv student-list.csv -i -o sorted-listing.csv

The -i option makes Eyegrade ignore students that are in the answers file but not in the student list. That is, the listing will only contain the students that are in the student list you provide.

This option may be useful, for example, if you examine students coming from different classes or groups. With this option you can produce a separate listing for each class.

4   Editing exams

Although you can use any software of your preference to typeset the exams, Eyegrade provides a module for doing that in combination to the LaTeX document preparation system.

First, write your questions in an XML document like the following one:

<?xml version="1.0" encoding="UTF-8"?>

<exam xmlns="http://www.it.uc3m.es/jaf/eyegrade/ns/"
      xmlns:eye="http://www.it.uc3m.es/jaf/eyegrade/ns/">

  <subject>Computers and More</subject>
  <degree>An Awesome Degree</degree>
  <title>Final examination</title>
  <date>January 1st, 2011</date>
  <duration>10 min.</duration>

  <question>
    <text>
      What is Python?
    </text>
    <choices>
      <correct>A programming language</correct>
      <incorrect>A computer manufacturer</incorrect>
      <incorrect>A kind of tree</incorrect>
      <incorrect>Who knows!</incorrect>
    </choices>
  </question>

  <question>
    <text>
      What is a webcam?
    </text>
    <choices>
      <correct>A video camera whose output may be viewed in real
         time over a network, especially over the Internet.</correct>
      <incorrect>A video camera intended to spy on spiders</incorrect>
      <incorrect>A kind of fish</incorrect>
      <incorrect>Who knows!</incorrect>
    </choices>
  </question>

  <question>
    <text>
      What is the thing at the right?
    </text>
    <code eye:position="right" eye:width="0.4">for letter in ['a', 'b', 'c']:
    print letter</code>
    <choices>
      <correct>A computer program</correct>
      <incorrect>A recipe</incorrect>
      <incorrect>A tree</incorrect>
      <incorrect>Who knows!</incorrect>
    </choices>
  </question>

  <question>
    <text>
      Is the thing in the right a logo?
    </text>
    <figure eye:width="0.16" eye:position="right">sample-logo.eps</figure>
    <choices>
      <correct>Well, it tries to be a logo, to be honest.</correct>
      <incorrect>No, it's a tree.</incorrect>
      <incorrect>No, it's a perfect square.</incorrect>
      <incorrect>Who knows!</incorrect>
    </choices>
  </question>

  <question>
    <text>
      Which is the capital of Spain?
    </text>
    <choices>
      <correct>Madrid.</correct>
      <incorrect>Barcelona.</incorrect>
      <incorrect>Paris.</incorrect>
      <incorrect>None of the other answers is correct.</incorrect>
    </choices>
  </question>

  <question>
    <text>
      Which is the capital of France?
    </text>
    <choices>
      <correct>Paris.</correct>
      <incorrect>Lion.</incorrect>
      <incorrect>Madrid.</incorrect>
      <incorrect>None of the other answers is correct.</incorrect>
    </choices>
  </question>

  <question>
    <text>
      What does ``hola'' mean in Spanish?
    </text>
    <choices>
      <correct>``Hello''.</correct>
      <incorrect>``Bye''.</incorrect>
      <incorrect>``Wave''.</incorrect>
      <incorrect>``Hola'' is not a word in Spanish.</incorrect>
    </choices>
  </question>

  <question>
    <text>
      How many months has a year?
    </text>
    <choices>
      <correct>12.</correct>
      <incorrect>6.</incorrect>
      <incorrect>10.</incorrect>
      <incorrect>It depends on whether it is a leap year or not.</incorrect>
    </choices>
  </question>

  <question>
    <text>
      Are dolphins mammals?
    </text>
    <choices>
      <correct>Yes.</correct>
      <incorrect>No.</incorrect>
      <incorrect>Sometimes.</incorrect>
      <incorrect>Who knows\ldots.</incorrect>
    </choices>
  </question>

  <question>
    <text>
      Who created Python?
    </text>
    <choices>
      <correct>Guido van Rossum.</correct>
      <incorrect>Richard Stallman.</incorrect>
      <incorrect>Dennis M. Ritchie.</incorrect>
      <incorrect>Jamie Zawinski.</incorrect>
    </choices>
  </question>

  <question>
    <text>
      How is ``Germany'' said in German?
    </text>
    <choices>
      <correct>Deutschland.</correct>
      <incorrect>Die Schweiz.</incorrect>
      <incorrect>Germany.</incorrect>
      <incorrect>Berlin.</incorrect>
    </choices>
  </question>

  <question>
    <text>
      What is Linux?
    </text>
    <choices>
      <correct>A free-software operating system.</correct>
      <incorrect>A river in Spain.</incorrect>
      <incorrect>The tallest mountain on Earth.</incorrect>
      <incorrect>An Internet company that sells phones.</incorrect>
    </choices>
  </question>

</exam>

Then, create a LaTeX template for the exam. This is an example:

\documentclass[a4paper,11pt]{article}

\usepackage{amssymb, amsmath}
\usepackage[utf8]{inputenc}
\usepackage{indentfirst}
\usepackage{multirow}

\usepackage[a4paper, margin=2.5cm, top=2cm, bottom=3cm]{geometry}

\pagestyle{empty}

{{declarations}}

\begin{document}

\begin{center}
\begin{tabular}{p{5.2cm}r}
\multirow{5}{*}[0.35cm]{\scalebox{0.18}{\includegraphics{sample-logo.eps}}} &
\Large  \textbf{{{subject}}} \\
& \textbf{{{degree}}} \\
& \\
{{date}} & {{title}}. \\
Duration: {{duration}}
      & Score: 5 points out of 10 total points for the exam. \\

\end{tabular}
\end{center}

\vspace{0.5cm}

\emph{There is only one correct answer for each multiple choice
  question.  Each correct answer adds 1 point.  Each incorrect answer
  has a penalty of $1/3$ points.  If no answer no score is awarded,
  neither positive nor negative.}

\vspace{0.5cm}

\begin{center}
\begin{tabular}{|p{0.8\textwidth}|}
\hline
\begin{itemize}
\item Mark out your answers with an ``X''.
\item If more than one answer or no answer are selected, the question
  is considered to be not answered (no score is awarded, neither
  positive nor negative).
\item Remember to fill in your name and student ID.
\end{itemize}
\\
\hline
\end{tabular}
\end{center}

\vspace{0.2cm}

\begin{center}
(1) Write your personal data clearly.
\end{center}

\begin{center}
\large

\begin{tabular}{|l|p{12cm}|}
\hline
Last name:   &  \\
\hline
First name: &    \\
\hline
Group:   &  \\
\hline
\end{tabular}
\end{center}

\vspace{0.2cm}

\begin{center}
(2) Write down your student ID and cross out your answers with an ``X'':
\end{center}

\begin{center}
\large
\textbf{Model:} {{model}}
\end{center}

{{id-box(9,ID)}}

{{answer-table}}

\clearpage

{{questions}}

\end{document}

In the template, notice that there are some marks within {{ and }} that are intended to be replaced by the script with data from the exam:

Note that a template is highly reusable for different exams and subjects.

Once the exam file and the template have been created, the script create_exam.py parses them and generates the exam in LaTeX format:

python -m eyegrade.create_exam -e exam-questions.xml -m 0AB template.tex -o exam

The previous command will create models 0, A and B of the exam with names exam-0.tex, exam-A.tex and exam-B.tex. Exam model 0 is a special exam in which questions are not reordered. The correct answer is always the first choice. Those files can be compiled with LaTeX to obtain a PDF that can be printed. In addition, the exam.eye file needed to grade the exam is automatically created (or updated if it already exists).

The script create_exam.py has other features, like creating just the front page of the exam (no questions needed). They can be explored with the command-line help of the program:

python -m eyegrade.create_exam -h

The answer table can be enlarged or reduced with respect to its default size, using the -S option and passing a scale factor (between 0.1 and 1.0 to reduce it, or greater than 1.0 to enlarge it). The following command enlarges the default size in a 50% (factor 1.5):

python -m eyegrade.create_exam -e exam-questions.xml -m A template.tex -o exam -S 1.5

5   Advanced features

5.1   Webcam selection

If your computer has more than one camera (e.g. the internal camera of the laptop and an external camera you use to grade the exams), Eyegrade will select one of them by default. If the selected camera is not the camera you want to use to grade the exams, use the -c <camera-number> option when invoking Eyegrade. Cameras are numbered 0, 1, 2, 3, etc. Invoke Eyegrade with a different camera number until the interface displays the one you want. For example, to select the camera numbered as 2:

python -m eyegrade.eyegrade exam.eye -c 2 -l student-list.csv

When the number is -1, eyegrade will automatically test different camera numbers until it finds one that works. When you select a camera number that does not exist or does not work, Eyegrade will also look automatically for other camera that works.

You can configure Eyegrade to always use a specific camera number by inserting the option camera-dev in the default section of the configuration file:

## Sample configuration file. Save it as $HOME/.eyegrade.cfg
[default]

## Default camera device to use (int); -1 for automatic selection.
camera-dev: 1

Save it in your user account with name .eyegrade.cfg. In Windows systems, your account is at C:\Documents and Settings\<your_user_name>.

5.2   Mixing manually-graded questions

You may want to mix in the same exam MCQ questions with other type of questions that must be graded manually. Even though Eyegrade can only grade the MCQ questions of the exam, it can simplify a little bit the process of mixing grades.

First, grade the MCQ exams with Eyegrade. Then, grade the other questions without changing the ordering of the exams.

Create a new CSV file with only one column, which contains the student ids of the students that submitted the exam. It will help a lot producing this listing in the same order you have graded the exams. Such a listing can be trivially obtained from the file eyegrade-answers.csv. In Linux, it can be done with just a command:

cut eyegrade-answers.csv -f 2 >extra-marks.csv

Edit that listing to include the marks of the manually-graded questions. Write marks in one or more columns at the right of the student id. Having this file the same order of your exams, introducing manual marks should be easier, since you do not need to search. This is an example with only one manual mark per exam (just one column):

100999991   7
100999997   8
100800003   5
100777777   9.5

The final listing that combines the results of all the questions can be produced with mix_grades:

python -m eyegrade.mix_grades eyegrade-answers.csv student-list.csv -x extra-marks.csv -o sorted-listing.csv

The columns with the manual marks would appear at the right in the resulting file:

100000333
100777777    7      13      9.5
100999997    15     1       8
100999991    9      6       7
100800003    7      13      5

5.3   Creating the exams in a word processor

The current prototype of Eyegrade require users to know LaTex in order to personalize exam templates. This section explains an alternative way to create exams compatible with Eyegrade in a word processor such as Microsoft Word. If you create your own exams with a word processor, you'll need also to edit the .eye file manually. See Manually editing the .eye file.

The objective is emulating the tables that Eyegrade creates so that the program can read them. This is an example:

Example answer tables.

You can use as a template this example MS Word document. It shows an answer table for 20 questions, which you can edit in order to customize if for your needs. Nevertheless, you should read the rest of this section if you are planning to customize the answer table.

An answer table is a table in which rows represent the questions and columns represent the choices. There can be more than one answer table, but they have to be side by side (they cannot be placed one above the other). The example above show two answer tables. A few restrictions have to be taken into account:

  • If there are more than one table, they must be horizontally aligned. That is, their top and bottom must be in the same line, and their rows must have exactly the same height (see the example above).
  • All the rows should have the same height.
  • In order to improve the detection process, the length of the vertical lines and the length of the horizontal lines should be more or less proportionate (e.g. one of them should not be more than a 30% larger than the other). If there are more than one answer table, consider the added length of the horizontal lines of every table. The following image illustrates this. The red vertical line is not much smaller than the sum of the two horizontal lines.
Example answer tables.
  • If an answer table has less rows than the others, it is better to keep the horizontal lines, as shown in the image below:
Example answer tables.

The boxes for the student ID number should be above the answer tables, not too close but not too far away either (see the example below). The width of the student ID table should be comparable to the sum of the width of the answer tables (approximately no less than 2/3 of that sum, and no more than 3/2). Student IDs with just a few digits (two, three, four) can potentially be problematic for wide answer tables.

Example answer tables with student ID box.

At the bottom of the answer boxes there must be some black squares. They encode the exam model (permutation). In addition, they help the system to know whether the detection of the answer tables was correct.

Imagine that there are two more rows at the end of each answer table, with the same height as the other rows. Squares will be either in the one above or in the one below, and there must be a square per column. Squares should be centered in those imaginary cells. The position (above/below) of a square conveys the information read by Eyegrade as binary information.

The exam model is encoded with three squares. Therefore, there can be eight different models. The fourth square is a redundancy code for the previous three squares. This 4-square pattern is repeated from left to right as long as there are columns. The table to which a column belongs is not taken into account. For example, if there are two answer tables with three columns each, the fourth square (the redundancy square) is placed at the first column of the second table. The other two columns of the second answer table would contain the same squares as the first two columns of the first table.

The following table show the 4-square pattern for each exam model, as they should be placed from left to right:

Model  
A Down Down Down Up
B Up Down Down Down
C Down Up Down Down
D Up Up Down Up
E Down Down Up Down
F Up Down Up Up
G Down Up Up Up
H Up Up Up Down

5.4   Manually editing the .eye file

The files that store the configuration of an exam and the correct answer for each question are stored with a .eye extension. An example is shown below:

[exam]
dimensions: 4,6;4,6
id-num-digits: 9

[solutions]
model-A: 4/1/2/1/1/1/2/4/1/2/3/1

[permutations]
permutations-A: 11{4,3,2,1}/12{1,3,4,2}/5{4,1,3,2}/3{1,4,3,2}/2{1,3,2,4}/9{1,4,3,2}/10{3,1,4,2}/1{3,2,4,1}/8{1,4,2,3}/6{2,1,3,4}/4{3,2,1,4}/7{1,3,2,4}

The file is just plain text and can be edited with any text editor. It has several sections: exam, solutions and permutations.

The fields of the exam section are:

  • dimensions: here the number of answer tables and the number of columns and rows in each answer table are configured. For example, "4,6;4,6" means that there are two answer tables, both of them with geometry "4,6". The "4" is the number of columns of the table. The "6" is the number of rows. Tables are specified from left to right (i.e. the first table geometry corresponds to the left-most table in the exam).
  • id-num-digits: number of cells of the table for the student id number. Putting a 0 here means that the id number needs not to be read.
  • correct-weight: a number, such as 1.75, that represents the score assigned to a correct answer.
  • incorrect-weight: a number that represents the score to be substracted for failed answers. Blank answers are not affected by this.

The fields correct-weight and incorrect-weight are optional. If they appear in the file, the program will show the total score in the user interface.

The solutions section specifies the correct answers for each model (permutation) of the exam. Models are identified by letters ("A", "B", etc.). For example:

model-A: 4/1/2/1/1/1/2/4/1/2/3/1
model-B: 3/2/1/4/4/2/2/1/4/2/3/3

In the example above, in the model A, the correct answer for the first question is the 4th choice, for the second question is the 1st choice, for the third question is the 2nd choice, etc.

The permutations section has information that allows to know how questions and choices have been shuffled with respect to the original order. They are used only for extracting statistics or fixing grades after the exam if the solutions used for grading are found to have an error in some questions. If you create the .eye manually, you probably want to just remove this section from the file, unless you need some of the above-mentioned functions.

5.5   Automatic detection of exam removal

If the camera in your setup is fixed, that is, you place an exam below the camera, review it, remove it and place the next exam, you may want Eyegrade to detect that you have removed the exam instead of having to click on the Save and capture next exam command.

You can activate this experimental feature in the Tools menu, Experimental submenu, option Continue on exam removal. When this option is checked, Eyegrade saves the current capture and enters the search mode automatically, after a few seconds of not detecting an exam. Before placing the new exam, wait for the system to actually enter the search mode: if you are too quick, Eyegrade might not detect the removal of the exam.

Tip: don't use this option if the camera is not fixed, because just moving it a little bit may cause Eyegrade to think that the exam has been removed.