Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Quantization handles conflicting labels incorrectly #609

Open
wants to merge 1 commit into
base: master
Choose a base branch
from

Commits on Nov 6, 2024

  1. Update quantization.cpp

    Fixes a bug where conflicting labels are handled incorrectly. 
    
    Minimal script for reproducing bug
    
    import MinkowskiEngine as ME
    import numpy as np
    #-----Dummy inputs
    #(Other examples without conflicting labels, and with one point at a time behaves as expected)
    #Case 0: 2 pairs of redundant points, 1) with same labels 2) with different labels.
    #Expected behaviour: coords_aug_new = [1, 0, 0], labels_new = [11, 255], feats = [0,0,101], return_index = [0, 2], return_mapping = [0, 0, ]
    #Result: one ignore label, but in the wrong position.
    coords_aug = np.float32(np.array([
    [1, 0, 0],[1, 0, 0],
    [2, 0, 0], [2, 0, 0]
    ]))
    labels = np.array([
    11,11,
    20,21,
    ])
    feats = np.array([
    [0, 0, 101],[0, 0, 101],
    [0, 0, 101], [0, 0, 101]
    ])
    return_args_tmp = ME.utils.sparse_quantize(
    coords_aug, feats, labels=labels.astype(np.int32), ignore_label=255, return_index = True, return_inverse = True), #quantization_size=1)
    (coords_aug_new,feats_new,labels_new,voxel_return_index, voxel_return_mapping)=return_args_tmp[0]
    print("New labels: ", labels_new)
    print("New coords: ", coords_aug_new)
    print("Return index", voxel_return_index)
    print("Return mapping", voxel_return_mapping)
    asbe authored Nov 6, 2024
    Configuration menu
    Copy the full SHA
    2262c4c View commit details
    Browse the repository at this point in the history