spine.model.singlep

Image classification module.

This module includes:

• Single full image classification

• Individual cluster classification

• UQ implementations of the full image classification

Classes

BayesianParticleClassifier(*args, **kwargs)
DUQParticleClassifier(*args, **kwargs)	Uncertainty Estimation Using a Single Deep Deterministic Neural Network https://arxiv.org/pdf/2003.02037.pdf Joost van Amersfoort, Lewis Smith, Yee Whye Teh, Yarin Gal.
EvidentialLearningLoss(*args, **kwargs)
EvidentialParticleClassifier(*args, **kwargs)
MultiLabelCrossEntropy(*args, **kwargs)
MultiParticleImageClassifier(*args, **kwargs)
MultiParticleTypeLoss(*args, **kwargs)

class spine.model.singlep.MultiParticleImageClassifier(*args: Any, **kwargs: Any)

Methods

__call__(*args, **kwargs)	Call self as a function.
forward(input[, clusts])	Run a batch of data through the forward function.
process_model_config(num_classes, **encoder)	Initialize the underlying encoder and the final layer.

split_input
split_input_as_tg_batch

MODULES = ['particle_image_classifier', 'network_base', 'mink_encoder']

split_input_as_tg_batch(point_cloud, clusts=None)

split_input(point_cloud, clusts=None)

forward(input, clusts=None)

Run a batch of data through the forward function.

Parameters:

data (TensorBatch) – (N, 1 + D + N_f) tensor of voxel/value pairs - N is the the total number of voxels in the image - 1 is the batch ID - D is the number of dimensions in the input image - N_f is the number of features per voxel

class spine.model.singlep.DUQParticleClassifier(*args: Any, **kwargs: Any)

Uncertainty Estimation Using a Single Deep Deterministic Neural Network https://arxiv.org/pdf/2003.02037.pdf Joost van Amersfoort, Lewis Smith, Yee Whye Teh, Yarin Gal.

Pytorch Implementation for SparseConvNets with MinkowskiEngine backend.

Methods

__call__(*args, **kwargs)	Call self as a function.
forward(input)	Run a batch of data through the forward function.
process_model_config(num_classes, **encoder)	Initialize the underlying encoder and the final layer.

bilinear
embed
update_buffers

MODULES = ['network_base', 'particle_image_classifier', 'mink_encoder']

embed(x)

bilinear(z)

forward(input)

Run a batch of data through the forward function.

Parameters:

data (TensorBatch) – (N, 1 + D + N_f) tensor of voxel/value pairs - N is the the total number of voxels in the image - 1 is the batch ID - D is the number of dimensions in the input image - N_f is the number of features per voxel

update_buffers()

class spine.model.singlep.EvidentialParticleClassifier(*args: Any, **kwargs: Any)

Methods

__call__(*args, **kwargs)	Call self as a function.
forward(input)	Run a batch of data through the forward function.
process_model_config(num_classes, **encoder)	Initialize the underlying encoder and the final layer.

MODULES = ['network_base', 'particle_image_classifier', 'mink_encoder']

forward(input)

Run a batch of data through the forward function.

Parameters:

data (TensorBatch) – (N, 1 + D + N_f) tensor of voxel/value pairs - N is the the total number of voxels in the image - 1 is the batch ID - D is the number of dimensions in the input image - N_f is the number of features per voxel

class spine.model.singlep.BayesianParticleClassifier(*args: Any, **kwargs: Any)

Methods

__call__(*args, **kwargs)

Call self as a function.

evidential_forward
forward
mc_forward
standard_forward

MODULES = ['network_base', 'mcdropout_encoder']

evidential_forward(input)

mc_forward(input, num_samples=None)

standard_forward(input, verbose=False)

forward(input)

class spine.model.singlep.MultiParticleTypeLoss(*args: Any, **kwargs: Any)

Methods

__call__(*args, **kwargs)

Call self as a function.

forward
forward_tg

forward_tg(out, valid_labels)

forward(out, type_labels)

class spine.model.singlep.MultiLabelCrossEntropy(*args: Any, **kwargs: Any)

Methods

__call__(*args, **kwargs)	Call self as a function.
calc_gradient_penalty(x, y_pred)	Code From the DUQ main Github Repository: https://github.com/y0ast/deterministic-uncertainty-quantification

forward

static calc_gradient_penalty(x, y_pred)

Code From the DUQ main Github Repository: https://github.com/y0ast/deterministic-uncertainty-quantification

Author: Joost van Amersfoort

forward(out, type_labels)

class spine.model.singlep.EvidentialLearningLoss(*args: Any, **kwargs: Any)

Methods

__call__(*args, **kwargs)

Call self as a function.

forward

forward(out, type_labels, iteration=0)