Question about model scaling

[Jiayuan-Gu]

Hi, thanks for the nice work. I am trying to rescale EGAD objects for grasping in the simulation. Indeed, object size (scale) is critical for grasping, as mentioned in IV.E of the paper.

I've seen a description of the scale used in the experiment.

Each object is scaled such that its minimum bounding box dimension is no more than 80% of the gripper width.

But according to the https://github.com/dougsm/egad/blob/master/scripts/prepare_meshes.py, the maximum dimension is first scaled to no more than the gripper width. It indicates chances are that a top-down grasping method can achieve high success rates regardless of shape complexity.

Can the authors clarify what kind of scaling strategies they actually use for the data generation and real-world experiments? Thanks.