spine.model.image

Whole-image classification/regression tasks.

Classes

ClusterImageClassifier(*args, **kwargs)
ImageClassLoss(*args, **kwargs)	Image classication loss.
ImageClassifier(*args, **kwargs)	Whole-image classification model.

class spine.model.image.ImageClassifier(*args: Any, **kwargs: Any)

Whole-image classification model.

This model uses various encoder declinations to classifier an entire image as belonging to a certain class.

model:
  name: image_class
  modules:
    classifier:
      # Image classifier configuration
    classifier_loss:
      # Image classifier loss configuration

Methods

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

MODULES = ['classifier', 'classifier_loss']

process_model_config(num_classes, **encoder)

Initialize the underlying encoder and the final layer.

Parameters:

• num_classes (int) – Number of classes that each image can be sorted as

• **encoder (dict) – Encoder configuration

forward(data)

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.image.ClusterImageClassifier(*args: Any, **kwargs: Any)

Methods

__call__(*args, **kwargs)

Call self as a function.

class spine.model.image.ImageClassLoss(*args: Any, **kwargs: Any)

Image classication loss.

Methods

__call__(*args, **kwargs)	Call self as a function.
forward(labels, logits, **kwargs)	Applies the image classification loss to a batch of data.
process_loss_config([loss, balance_loss, ...])	Initialize the loss function
process_model_config(num_classes, **encoder)	Initialize the underlying encoder and the final layer.

process_model_config(num_classes, **encoder)

Initialize the underlying encoder and the final layer.

Parameters:

• num_classes (int) – Number of classes that each image can be sorted as

• **encoder (dict) – Encoder configuration

process_loss_config(loss='ce', balance_loss=False, weights=None)

Initialize the loss function

Parameters:

• loss (str, default 'ce') – Name of the loss function to apply

• balance_loss (bool, default False) – Whether to weight the loss to account for class imbalance

• weights (list, optional) –

  3. One weight value per class

forward(labels, logits, **kwargs)

Applies the image classification loss to a batch of data.

Parameters:

• labels (List[int]) –

  2. List of image labels, one per entry in the batch

• logits (TensorBatch) – (B, C) Tensor of predicted logits, one per entry in the batch

• **kwargs (dict, optional) – Other labels/outputs of the model which are not relevant here

Returns:

• loss (torch.Tensor) – Value of the loss

• accuracy (float) – Value of the image-wise classification accuracy