GraspGen
Introduction
GraspGen (https://github.com/NVlabs/GraspGen/tree/main) is a machine learning framework that uses a diffusion-based model to generate 6-DOF (Degrees of Freedom) robotic grasp poses. In simple terms, it acts as a "brain" that analyzes a 3D object—whether from a complete mesh or a partial point cloud from a camera—and calculates the exact 3D coordinates and rotation angles needed for a robotic gripper to successfully grab it.It is designed to be fast (real-time prediction), efficient, and adaptable to single objects as well as cluttered environments.
Support
Robotiq 2F gripper is part of GraspGen embedded assest.