Synthetic dataset (R1) - We note that many of the (synthetic or real) datasets prepared in robotic grasp learning

We thank all reviewers for their constructive comments. Due to space limit, we address the major concerns as follows. This is the main reason why we have to prepare our own synthetic data. We will improve the description of GPNet-Naive in the paper. Our grasp parametrization using surface contact, grasp center, and'pitch' angle Fig.1), where we prune grasp proposals by antipodal We will include these results in the paper.