What do the rows and columns in object_taxonomies.npy represent ?

[daichi-3110]

Hello.

I would like to refer to the grasps that you have associated with the object, so I am processing object_taxonomies.npy.
This npy include 58x33 matrix. Maybe I think 58 is the number of objects included in YCB object set CAD models, and 33 is the number of grasps. Is it right?
What is the order? Are they lined up in numerical order? Please let me know.

Thank you very much.

[enriccorona]

Hi,
For each grasp we also annotated the grasp type (according to this paper https://www.is.mpg.de/uploads_file/attachment/attachment/256/grasp_taxonomy.pdf) and tried to have a diverse set of natural grasp for all objects.

There are 33 grasp types in total. As you said, 58 is the number of objects, and we identify as a one-hot encoding which of these 33 grasps are available. Note that the list is non-exclusive and there could be grasps from other grasp types

[daichi-3110]

Do the indices of the rows in this matrix correspond to the order in which the objects included in YCB object set CAD models are arranged in numerical order? For example, index 0 represents 002_master_chef_can, index 1 represents 003_cracker_box, index 2 represents 004_sugar_box, ..., index 57 represents 077_rubiks_cube.

Also, Do the indices of the columns in this matrix correspond to the assigned index of grasp in Fig.4 in the paper written by Feix et al. For example, index 0 represents Large Diameter, index 1 represents Small Diameter, index 2 represents Medium Wrap, ..., index 32 represents Inferior Pincer.