Dependencies

The code was written for Python 2.7 and requires the following additional packages:

• NumPy 1.7
• PyQt4 4.10.3
• boto 2.13.3
• Matplotlib 1.3.0

They can most easily installed using pip, or easy_install.

Initial set up

Boto

After having installed the dependencies, you need to create a .boto file in your home directory which contains the Amazon Web Service (AWS) credentials. The file must contain the following lines:

[Credentials]  
aws_access_key_id = YOURACCESSKEY  
aws_secret_access_key = YOURSECRETKEY  

See also Boto Tutorial

Dropbox

The code makes use of Dropbox to host the webinterface and data for the Turkers. So you need a Dropbox account and a folder "Public" in your Dropbox. The public folder can be created, if it doesn't exist yet.

Now you only have to provide the local path to your public folder and the url to your public folder in the [Image storage] section of the config files mturk_segmentation.ini and mturk_features.ini:

[Image storage]
host-url = URL
dropbox-path = LOCALPATH

Setting up a new MTurk annotation project

To start a new Mechanical Turk annotation task you need at least a VideoLabelME xml file. You need to set up the following folder structure:

• MTurkAnnotation
  • YourProjectName
    • Segmentation
      • YourProjectName
        • Project image files
      • VideoLabelME XML file
      • "mturk_segmentation.ini" config file
    • Features
      • YourProjectName
        • Project mages files
      • Turked VideoLabelME XML file
      • "mturk_features.ini" config file

The MTurk client is started from the command line, with the VideoLabelME XML file of your project as the only argument.

python mturkclient.py VideoLableME.xml

Here is a screenshot of the GUI:

images/mturkclient_gui.png