Return to the previous page. Paul Fearing's home page.

  Validation of snow-covered scenes is hard, in that snow observed outdoors is the result of uncontrollable and unknown environmental factors. Creating artificial snow is beyond our capabilities as a graphics lab, so instead we restrict validation to observation, asking the question: ``does our algorithm produce phenomena and/or effects that are observable in nature?'' 
However, we were able to perform a few simple experiments to show that our snow stability algorithms are at least plausible. We substituted sifted flour for snow, to improve controllability and show that our algorithms work for materials other than snow. The figure below shows a side-by-side comparison of real and computer generated scenes. 
A real flour covered scene (left) and a computer generated scene (right) compared to show that our stability algorithms are at least plausible. Our experimental setup was fairly ad-hoc: despite our best efforts, flour was distributed unevenly around the base of  the real sphere.
Click to enlarge.
The picture below compares the flake dusting on a real and computer generated sign. Note the dusting densities around the top and edges of each sign.
Comparison of flake dusting on a real sign (left) and a computer generated sign (right).
Click to enlarge.