Project #05
Title: Using Summingbird for aggregating eye tracking data to find patterns in images in a multi-user environment
Leader's Name: Johan Fogelström
Member2 Name: Remzi Can Aksoy
Related paper: Oscar Boykin, Sam Ritchie, Ian O’Connell, and Jimmy Lin. 2014. Summingbird: A Framework for Integrating Batch and Online MapReduce Computations. In Proceedings of the VLDB Endowment, Volume 7, 2013-2014.
Presentation Day: May 20
Model: ES
Abstract:
The aim of this project is to create a tool that using existing eye trackers can determine interesting areas of a large set of images.
The tool could be used to train machine learning algorithms with learning data produced by the tool in order to automatically evaluate images for relevancy to the audience.
To do this, lots of data is required, data which will be provided both by data streams and stored data from earlier experiments.
For this reason, we use a framework designed to handle both batch and on-line application of the MapReduce technique known as Summingbird, developed by researchers at Twitter to reduce complexity when developing algorithms that can handle their data sets, since the same algorithm would otherwise have to be developed twice to handle both cases.
An eye tracker is a device (or software using general purpose hardware) to gather data about a specific users eye movements and translate this into, among other things coordinates on the screen plane where the user is currently focused.
Eye trackers is good for providing this data because they give measurable data on what the subject instinctively react to in their field of vision.
Asking subjects through questionnaires or interviews could introduce bias, expectations and other data distorting errors.