==========================================
Mesh Saliency via Local Curvature Entropy
==========================================
EUROGRAPHICS 2016

Supplemental Material 
==========================================

These are the experimental results of saliency detection for this paper.
All saliency values have been normalized to a range [0,1].

* Directory "results-lce-*" contains results from our LCE method.

* Directories "results-spectral-3K-*" and "results-spectral-5K-*" contain
  results from the spectral method (Song et al., TOG 2014), using a setting of 
  3,000 and 5,000 faces for QSlim simplification

* Directory "results-spectral-5K-*" contains results from the spectral method
  (Song et al., TOG 2014), using a setting of 5,000 faces for QSlim simplification
  
* Directories ending with "heat" contain results using a "heat" color map
  (same as in the paper, see heat.png)
  
* Directories ending with "rainbow" contain results created using the
  classical "rainbow" color map (see rainbow.png)
  
* Color mapping was directly performed using the normalized values from [0,1]
  (i.e., there was no additional clipping of min/max values / percentile to a 
   more narrow range, for example)

* Each mesh has a custom, normalized float property entitled "quality".
  This property can be used to apply a custom color mapping for comparison.
  For example, to use Meshlab's inverse rainbow function, load the mesh and choose
  "Filters"->"Color Creation and Processing"->"Colorize by vertex Quality"