GUARD: Geometric Uncertainty-Aware Point Cloud Denoising and Segmentation for Robotic Hard Disk Drive Disassembly

GUARD is a geometric uncertainty-aware framework that improves point cloud segmentation for robotic disassembly by distinguishing between genuine component geometry and scanning artifacts. It combines a multi-scale geometric transformer with a Gaussian Process to estimate per-point uncertainty, achieving better performance than predictive entropy in detecting corrupted points.