Biomimetics
1 Sept 2015

Pick it with your eyes – an eye tracking experience

Eye-tracker.
Eye-tracker.

Automatic extraction of features from scientific images is important in the field of space exploration. During space exploration a large number of images are taken and analyzing all of them in real-time back on Earth is impossible due to delays in transmission. Moreover, it is unpractical to store all the images for an offline analysis. The ideal solution would be to evaluate in real-time the scientific content of a picture, based on relevance of the content. The development of machine learning tools could enable autonomous identification of interesting pictures, but the preparation of training sets is usually done manually, by labeling of interesting items on pictures.

Project overview

The goal of this small project is to label the pictures for a training set by using the gaze of humans recorded with an eye tracker. By training an algorithm on this dataset, a robot could learn to look at a scene in a similar way as humans. Below is an example set. On each picture a heat map marks the time participants were allocating to each area. Additionally, the pictures were marked in red by geologists based on presence of interesting features.

Outcome

Bio-Engineering Ariadna Final Report
Curiosity Cloning: Neural Modelling for Image Analysis
Healy, G. and Wilkins, P. and Smeaton, A.F. and Izzo, D. and Rucinski, M. and Ampatzis, C. and Moraud, E.M.
European Space Agency, the Advanced Concepts Team, Ariadna Final Report 08-8201b
(2010)
Download
BibTex
Hamburger icon
Menu
Advanced Concepts Team