Generalizable 6D Pose Estimation of Textureless Objects with Planar-based Gaussian Splatting
PG-Pose, a geometry-aware framework combining Planar-based Gaussian Splatting (PGS) reconstruction and Geometry-driven pose optimization for industrial robots in grasping tasks, is proposed and deployed on a dual-arm industrial robot and successfully realizes the grasping task on an unseen object.