Imagine that a natural history museum gives you a collection of triangle meshes acquired with a 3D laser scanner (e.g., scans of fossils). Each mesh in the collection is labeled to indicate in which of two categories it is a member (e.g., species A and specied B).

The curator of the museum then asks you to design a computer system that can identify the salient differences between meshes in the two categories. She wants the system to produce an accurate category prediction for new meshes based on similarities to ones in the collection and wants it to provide visualizations indicating which regions of the surface meshes are salient for distinguishing the categories.

Unfortunately, the differences in surface shapes between categories can be quite subtle, either because only small parts of the surface are different (e.g., a deviation in the shape of an eye socket), or because differences are due to long-range, low-frequency warps that are difficult to observe visually (e.g., small differences in the curvature of a skull).

a) Please describe the overall design of your proposed system, possibly with a flow-chart diagram outlining the flow of data through your system if it has multiple steps.

b) Please describe the 2-3 most important algorithms in your system. For each, please explain what subproblem it is addressing (e.g., alignment, feature detection, etc.), what algorithm you have chosen to address the subproblem, and why you think that algorithm is addressing the curator's problem.

Your entire answer should fit within 1-2 pages. Please save/scan your answers into a pdf file named written_exercise3.pdf and submit it to CS dropbox.