question about FOR-Instance dataset

[huiminw519-dotcom]

I'm trying to run the training script with your FOR-Instance dataset but encountering an error. I ran this command:

python train.py task=panoptic data=panoptic/treeins_set1_add_return_num models=panoptic/FORpartseg_3heads model_name=PointGroup-PAPER training=treeins_set1_return_num job_name=#YOUR_JOB_NAME#

And got this error:

ValueError: no field of name Sum

When I checked the PLY files in the dataset, I found they only contain these 6 fields: x, y, z (coordinates)、intensity、semantic_seg、treeID

But the code expects 17 additional geometric features including:

Sum, Omnivariance, Eigenentropy, Anisotropy

planarity, linearity, Surface_var, scattering

verticality, Verticality2

moment1_1, moment1_2, moment2_1, moment2_2

return_num, scan_angle_rank

Questions:

How many features does the FOR-Instance dataset actually contain?

Did I miss a preprocessing script? Is there a step to compute these geometric features from the raw point clouds that I should run first?

Thanks for your help!