mednet.models.segment.hed

Holistically-Nested Edge Detection (HED) network architecture, from [XIE-2015].

Classes

ConcatFuseBlock()

Take in five feature maps with one channel each, concatenates thems and applies a 1x1 convolution with 1 output channel.

HED([loss_type, loss_arguments, ...])

Holistically-Nested Edge Detection (HED) network architecture, from [XIE-2015].

HEDHead(in_channels_list)

HED head module.
class mednet.models.segment.hed.ConcatFuseBlock[source]

Bases: Module

Take in five feature maps with one channel each, concatenates thems and applies a 1x1 convolution with 1 output channel.

forward(x1, x2, x3, x4, x5)[source]

Define the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

class mednet.models.segment.hed.HEDHead(in_channels_list)[source]

Bases: Module

HED head module.

Parameters:

in_channels_list (list) – Number of channels for each feature map that is returned from backbone.

forward(x)[source]

Define the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.

class mednet.models.segment.hed.HED(loss_type=<class 'mednet.models.segment.losses.MultiSoftJaccardBCELogitsLoss'>, loss_arguments={}, optimizer_type=<class 'torch.optim.adam.Adam'>, optimizer_arguments={}, scheduler_type=None, scheduler_arguments={}, model_transforms=[], augmentation_transforms=[], num_classes=1, pretrained=False)[source]

Bases: Model

Holistically-Nested Edge Detection (HED) network architecture, from [XIE-2015].

Parameters:

  • loss_type (type[Module]) –

    The loss to be used for training and evaluation.

    Warning

    The loss should be set to always return batch averages (as opposed to the batch sum), as our logging system expects it so.

  • loss_arguments (dict[str, Any]) – Arguments to the loss.

  • optimizer_type (type[Optimizer]) – The type of optimizer to use for training.

  • optimizer_arguments (dict[str, Any]) – Arguments to the optimizer after params.

  • scheduler_type (type[LRScheduler] | None) – The type of scheduler to use for training.

  • scheduler_arguments (dict[str, Any]) – Arguments to the scheduler after params.

  • model_transforms (Sequence[Callable[[Tensor], Tensor]]) – An optional sequence of torch modules containing transforms to be applied on the input before it is fed into the network.

  • augmentation_transforms (Sequence[Callable[[Tensor], Tensor]]) – An optional sequence of torch modules containing transforms to be applied on the input before it is fed into the network.

  • num_classes (int) – Number of outputs (classes) for this model.

  • pretrained (bool) – If True, will use VGG16 pretrained weights.

forward(x)[source]

Same as torch.nn.Module.forward().

Parameters:

  • *args – Whatever you decide to pass into the forward method.

  • **kwargs – Keyword arguments are also possible.

Returns:

Your model’s output

set_normalizer(dataloader)[source]

Initialize the normalizer for the current model.

This function is NOOP if pretrained = True (normalizer set to imagenet weights, during contruction).

Parameters:

dataloader (DataLoader) – A torch Dataloader from which to compute the mean and std. Will not be used if the model is pretrained.

Return type:

None