GraspNet

Table of contents

1. Core ideas
2. Typical workflow
3. Manipulation relevance

---

GraspNet is a large-scale 6-DoF grasp detection framework that predicts diverse grasp poses from point-cloud observations of cluttered scenes.

Core ideas

• Dense grasp proposals over observed 3D geometry
• Grasp quality scoring for ranking candidates
• Dataset and benchmark support for large-scale evaluation

Typical workflow

1. Acquire an RGB-D observation and build a point cloud.
2. Generate candidate 6-DoF grasps on the scene.
3. Score and filter grasps by quality and collision checks.
4. Select and execute a high-quality feasible grasp.

Manipulation relevance

• Strong baseline for cluttered-scene 6-DoF grasping
• Useful for comparing learned grasp detectors
• Connects perception outputs to executable grasp poses