PDF version of this entire document

Goals

Outline of the work and general thoughts about the goal suggest an aspiration towards GMDS vs. PCA comparisons, whereby shrewd parallelisation4 of the former method may deem it more suitable for many practical needs. At the time of writing we approach this goal by writing code that can handle the data by normalising it somewhat before using the ICP algorithm to process pairs and build models separately. We also work on several different visualisation methods, cleanup methods, smoothing algorithms, and cropping methods. We are classifying images and expressions as neutral and not neutral by dividing the datasets prior to runtime, rather than determine this 'on the fly'.



Roy Schestowitz 2012-01-08