Learning robotic grasps from visual observations is a promising yet challenging task. Recent research shows its great potential by preparing and learning from large-scale synthetic datasets. For the popular, 6 degree-of-freedom (6-DOF) grasp setting of parallel-jaw gripper, most of existing methods take the strategy of heuristically sampling grasp candidates and then evaluating them using learned scoring functions. This strategy is limited in terms of the conflict between sampling efficiency and coverage of optimal grasps. To this end, we propose in this work a novel, end-to-end \emph{Grasp Proposal Network (GPNet)}, to predict a diverse set of 6-DOF grasps for an unseen object observed from a single and unknown camera view. GPNet builds on a key design of grasp proposal module that defines \emph{anchors of grasp centers} at discrete but regular 3D grid corners, which is flexible to support either more precise or more diverse grasp predictions. To test GPNet, we contribute a synthetic dataset of 6-DOF object grasps; evaluation is conducted using rule-based criteria, simulation test, and real test. Comparative results show the advantage of our methods over existing ones. Notably, GPNet gains better simulation results via the specified coverage, which helps achieve a ready translation in real test.

Additional Feedback: - The advantage of this approach over [33][34] as mentioned in Line 40 is mostly computational. However, no computational analysis is done to support this claim. Do these approaches achieve a diverse set of robust grasps when given enough time - how much time does it take. The code for these approaches is publicly available. Is there a theoretical limitation of the approach?

This paper proposes an approach to predict multiple stable 6-dof grasp parameters for standard parallel-jaw grippers from object point cloud inputs, with associated confidence values. Grasps are represented as tuples of (contact points of the 2 jaws and the pitch angle of the gripper), which motivates the new architectural choices proposed here, inspired by standard architectures in 2D object detection. While the network is trained end-to-end, it is internally decomposed in a sensible stage-wise manner. They also create a synthetic 22.6M 6-DOF grasp dataset built on ShapeNet objects using physics simulation, which upon public release, will be the largest such dataset. Finally, there are some limited transfer results that demonstrate transferability to real-world grasping with acceptable performance drop.

Learning robotic grasps from visual observations is a promising yet challenging task. Recent research shows its great potential by preparing and learning from large-scale synthetic datasets. For the popular, 6 degree-of-freedom (6-DOF) grasp setting of parallel-jaw gripper, most of existing methods take the strategy of heuristically sampling grasp candidates and then evaluating them using learned scoring functions. This strategy is limited in terms of the conflict between sampling efficiency and coverage of optimal grasps. To this end, we propose in this work a novel, end-to-end \emph{Grasp Proposal Network (GPNet)}, to predict a diverse set of 6-DOF grasps for an unseen object observed from a single and unknown camera view.