GCNGrasp-VP: Affordance-Guided View Planning for Efficient Task-Oriented Grasping
Paper • 2606.19091 • Published
GCNGrasp-v2
GCNGrasp-v2 is a task-oriented grasp (TOG) model that simultaneously performs grasp evaluation and affordance field prediction with constant-time inference complexity $\mathcal{O}(1)$. It is the core component of the GCNGrasp-VP framework, enabling efficient robotic grasping under occlusions through affordance-guided view planning.