shuffle makes no sense

[zhixiongzh]

Hi,
in the file function voxelize_spconv, which is in model_zoom.py, you shuffle all points by
cat_pt_ind = cat_pt_ind[shuffled_ind,:]

and then the output unq and processed_pooled_data is still in the order of the grid, which is done automatically by scatter_max function.
pooled_data = torch_scatter.scatter_max(processed_cat_pt_fea, unq_inv, dim=0)[0]

It seems this step makes no point in the function.

Usually, we shuffle the data in the dataloader, which has been done automatically, instead of shuffle the date within the batch.

[hongfz16]

Hi,

Thanks for the question.

Actually, the shuffling operation here is preparing the data for the random sampling in this line. If we do not perform shuffling, then the sampling will not be random.

As for the scatter operation later, it has no relationship with the shuffling operation above. The above shuffling is done at the point level. Later the randomly sampled points will be converted to voxel by scatter operation.

If you have further confusion, feel free to continue this discussion.