Measuring 3D Mesh Saliency using an Eye Tracker

Thomas Sweetman


Supervised by Yukun Lai; Moderated by Paul L Rosin

In recent years 3D data has become popular as low-cost devices such as Kinect have emerged. There are many applications of 3D data, such as gaming, cinematography special effects, 3D printing, etc. When these models are manipulated (eg downsized, stretched, morphed, etc) then important (ie salient) features should be preserved as best as possible.

This project will develop a method for measuring the salience of parts of the 3D mesh data. Experiments will be carried out with an eye tracker to provide an indication of saliency that will be consistent with human judgement Existing models are based on low level geometric features extracted from the data Using machine learning and the eye tracking data, the student will optimise such a model to better predict human perception of salience.

Initial Plan (05/02/2018) [Zip Archive]

Final Report (11/05/2018) [Zip Archive]

Publication Form