2012 Winter Project Week:PNSnormals
Principal Nested Spheres methodology
Key Investigators
- UNC: Beatriz Paniagua, Martin Styner, Steve Pizer
- University of Pittsburg: Sungkyu Jung
- Utah SCI: Josh Cates, Manasi Datar, Ross Whitaker
Objective
Compute normal discrepancies using Principal Nested Spheres (PNS) in entropy-based particle systems. In the proposed methods normals projected into the unit sphere, then we fit the great circle that approximates the data and the Frechet mean of that data, that lives in the great circle. Residuals are computed by how much the individual data differs from the Frechet mean. Our idea is use this normal information (residuals) as attribute data and incorporate it to the global optimization of the system. In this way, normals are not handled as a penalty.
Approach, Plan
The code is up and running but still does not provide the desired results. Meeting with the SCI crew to show preliminary data and troubleshoot possible misconceptions and errors, produced maybe during the integration of our PNS code with ShapeWorks. Revise battery of tests already computed by Jan 2012. Decide next steps.
Progress
A battery of tests was created for debugging purposes this week. After several discussions with the Utah group, we discovered potential conceptual errors and ways to finish the implementation and start testing.
Delivery Mechanism
This work will be delivered in NITRC, as an extension to the current ShapeWorks pipeline
