Bartertown: a Single-Player Human Computation Game to Create a Dataset of Iconic Gestures
Wouter van den Heuvel
This work explores the use of the human computation games paradigm to create a dataset of iconic gestures describing primitive 3D shapes. To this end we have created Bartertown, a single-player video game using Microsoft Kinect v2. Contrary to many human computation games, our game is not just about providing annotations to existing data, instead players are the primary data providers. Data quality is achieved by virtue of a self validating system. We assert that the medium of games can be employed successfully in these scenarios. We reflect on the prototype and propose a list of general guidelines useful for researchers interested in creating human computation games for acquisition of gesture corpora.
Wouter van den Heuvel, "Bartertown: a Single-Player Human Computation Game to Create a Dataset of Iconic Gestures", Master's Thesis for the Media Technology programme, Leiden University (The Netherlands), 2015