Problem
Point-cloud classifiers are trained on complete shapes, but real scans are partial.
What I did
Phase 1: benchmark how missing geometry (random drop, patch and part removal, planar cut, limited views) changes classifier accuracy and calibration.
Role
Phase 1 of a team project (CS 6501 3D Computer Vision): a benchmark of how missing geometry affects point-cloud classifiers.
Credits
Team project.
Links
No public links yet.