Surface Reconstruction with Neural Meshes.

Ivrissimtzis, Ioannis; Jeong, Won-Ki; Lee, Seungyong; Lee, Yunjin Lee; Seidel, Hans-Peter

Abstract

We describe a Learning algorithm for surface reconstruction based on an incrementally expanding Neural Network known as Growing Cell Structure. The Neural Network (Neural Mesh) expands by splitting its most active vertices and removing its least active. It learns topologies by removing triangles to create boundaries and merging boundaries to create handles. High-level ensemble and forgetting operations are used to enhance the algorithm’s performance. We study the algorithm theoretically, analyzing statistically its main components, and experimentally, using an extensive range of input data sets with varying amounts of pre-processing.

Citation

Ivrissimtzis, I., Jeong, W.-K., Lee, S., Lee, Y. L., & Seidel, H.-P. (2005, May). Surface Reconstruction with Neural Meshes. Presented at Mathematical Methods for Curves and Surfaces., Tromsoe, Norway

• Presentation Conference Type: Conference Paper (published)
• Conference Name: Mathematical Methods for Curves and Surfaces.
• Publication Date: 2005-05
• Pages: 223-242
• Public URL: https://durham-repository.worktribe.com/output/1697755