[Bf-committers] improved UV unwrapping need skilled math folks (dimensionality reduction)

[Tom M]

any smart cookies with advanced mathematics interested in tackling an
algorithm that appears to give much higher quality than ABF++ for UV
unwrapping?

Apparently using ISOMAP can give a drastic reduction in stretching,

http://www.polygonal-design.fr/e_unfold/gen71.php

and there are more advanced dimensional reduction techniques
(including improvements on ISOMAP) that should give even better
results.

http://en.wikipedia.org/wiki/Nonlinear_dimensionality_reduction
http://vodinhphong.wordpress.com/2010/12/14/resources-on-manifold-learning-1/

It would be good to reclaim our crown as providing the best quality
initial unwrap :)

As an asside, we should really consider seeking out some computer
vision/computer learning/statistics folks since they are heavily into
dimensionality reduction.

LetterRip