About CloudCrop #46
Description
Dear author, I have noticed that in the CloudCrop module, you have extracted point features from the cylindrical area with a determined approach direction in the following way：
grouped_feature = self.grouper(seed_xyz_graspable, seed_xyz_graspable, vp_rot,seed_features_graspable)
but in graspnet，same modules use：
grouped_features.append(grouper( pointcloud, seed_xyz, vp_rot
I would like to ask, shouldn't the features be statistically analyzed from the original point cloud? If the original point is set as the capture point, it means that there is only one capture point in each group. So, what's the use of grouping if this is done?