Boundary Loss proposed in: Alexey Bokhovkin et al., Boundary Loss for Remote Sensing Imagery Semantic Segmentation https://arxiv.

This is the combination of Cross Entropy and Dice Loss.

Class that contains LaserScan with x, y, z, r.

Class that contains LaserScan with x, y, z, r, sem_label, sem_color_label, inst_label, inst_color_label.

Dice loss of binary class Args: smooth: A float number to smooth loss, and avoid NaN error, default: 1 p: Denominator value: \sum{

Dice loss of binary class Args: smooth: A float number to smooth loss, and avoid NaN error, default: 1 p: Denominator value: \sum{

This criterion is based on Dice coefficients. Modified version of: https://github.com/ai-med/nn-common-modules/blob/master/nn_common_modules

This criterion is based on Dice coefficients. Modified version of: https://github.com/ai-med/nn-common-modules/blob/master/nn_common_modules

Dice loss, need one hot encode input Args: weight: An array of shape [num_classes,] ignore_index: class index to ignore pr

Dice loss, need one hot encode input Args: weight: An array of shape [num_classes,] ignore_index: class index to ignore pr

This uses a different encoding from v1. v1: [cls1, cls2, ..., other1_for_group0, other_for_group_1, bg, bg_others] this: [group0_others,

This uses a different encoding from v1. v1: [cls1, cls2, ..., other1_for_group0, other_for_group_1, bg, bg_others] this: [group0_others,

r"""Online visualizer implemented with Open3d. Args: points (numpy.array, shape=[N, 3+C]): Points to visualize. The Points clo

r"""Online visualizer implemented with Open3d. Args: points (numpy.array, shape=[N, 3+C]): Points to visualize. The Points clo

Inference dataset for loading an unpacked sequence of Waymo

Computes and stores the average and current value

Boundary Loss proposed in: Alexey Bokhovkin et al., Boundary Loss for Remote Sensing Imagery Semantic Segmentation https://arxiv.

This is the combination of Cross Entropy and Dice Loss.

Refer to https://github.com/tianweiy/CenterPoint

Local Geometric Anisotropic

This is the autograd version, you can also try the LabelSmoothSoftmaxCEV2 that uses derived gradients

Basic wrapper around `opt` to simplify hyper-parameters changes.