Net2VecLoss
- class hybrid_learning.concepts.train_eval.kpis.batch_kpis.Net2VecLoss(factor_pos_class=0.5, reduction=BatchReduction.mean, target_thresh=0.0)[source]
Bases:
AbstractIoULossSimplified intersection over union as loss. This loss is the one used for the original implementation of the Net2Vec framework [Fong2018] (even though this is a rewrite and no code is used from there). It works on masks of prediction and ground truth (gt) indicating the foreground (fg) area. The masks may be binary, non-binary or mixed. The target masks are binarized.
Given For an instance, it calculates as
(1)\[\text{Net2Vec}(instance) = b \cdot TP + (1-b) \cdot TN\]with
TP: true positives, resp. the intersection of predicted fg area and gt fg area
TN: true negatives, resp. the intersection of predicted background (bg) area and gt bg area
The following tensor dimensions are allowed:
1D: The tensor is assumed to be 1D without batch dimension.
2D: The tensor is assumed to be 2D without batch dimension.
>2D: The tensor is assumed to be 2D with batch dimension 0, width dim. -1, height dim. -2.
- Fong2018
R. Fong and A. Vedaldi, “Net2Vec: Quantifying and explaining how concepts are encoded by filters in deep neural networks” in Proc. 2018 IEEE Conf. Comput. Vision and Pattern Recognition, Salt Lake City, UT, USA, 2018, pp. 8730–8738, https://arxiv.org/abs/1801.03454
Public Data Attributes:
Settings to reproduce the instance.
Inherited from : py: class:AbstractIoULoss
Settings to reproduce the instance.
Public Methods:
forward(outputs, targets)Calculate Net2Vec loss (1).
Inherited from : py: class:Module
forward(outputs, targets)Calculate Net2Vec loss (1).
register_buffer(name, tensor[, persistent])Adds a buffer to the module.
register_parameter(name, param)Adds a parameter to the module.
add_module(name, module)Adds a child module to the current module.
get_submodule(target)Returns the submodule given by
targetif it exists, otherwise throws an error.get_parameter(target)Returns the parameter given by
targetif it exists, otherwise throws an error.get_buffer(target)Returns the buffer given by
targetif it exists, otherwise throws an error.apply(fn)Applies
fnrecursively to every submodule (as returned by.children()) as well as self.cuda([device])Moves all model parameters and buffers to the GPU.
xpu([device])Moves all model parameters and buffers to the XPU.
cpu()Moves all model parameters and buffers to the CPU.
type(dst_type)Casts all parameters and buffers to
dst_type.float()Casts all floating point parameters and buffers to
floatdatatype.double()Casts all floating point parameters and buffers to
doubledatatype.half()Casts all floating point parameters and buffers to
halfdatatype.bfloat16()Casts all floating point parameters and buffers to
bfloat16datatype.to_empty(*, device)Moves the parameters and buffers to the specified device without copying storage.
to(*args, **kwargs)Moves and/or casts the parameters and buffers.
register_backward_hook(hook)Registers a backward hook on the module.
register_full_backward_hook(hook)Registers a backward hook on the module.
register_forward_pre_hook(hook)Registers a forward pre-hook on the module.
register_forward_hook(hook)Registers a forward hook on the module.
state_dict([destination, prefix, keep_vars])Returns a dictionary containing a whole state of the module.
load_state_dict(state_dict[, strict])Copies parameters and buffers from
state_dictinto this module and its descendants.parameters([recurse])Returns an iterator over module parameters.
named_parameters([prefix, recurse])Returns an iterator over module parameters, yielding both the name of the parameter as well as the parameter itself.
buffers([recurse])Returns an iterator over module buffers.
named_buffers([prefix, recurse])Returns an iterator over module buffers, yielding both the name of the buffer as well as the buffer itself.
children()Returns an iterator over immediate children modules.
named_children()Returns an iterator over immediate children modules, yielding both the name of the module as well as the module itself.
modules()Returns an iterator over all modules in the network.
named_modules([memo, prefix, remove_duplicate])Returns an iterator over all modules in the network, yielding both the name of the module as well as the module itself.
train([mode])Sets the module in training mode.
eval()Sets the module in evaluation mode.
requires_grad_([requires_grad])Change if autograd should record operations on parameters in this module.
zero_grad([set_to_none])Sets gradients of all model parameters to zero.
share_memory()extra_repr()Set the extra representation of the module
Special Methods:
__init__([factor_pos_class, reduction, ...])Init.
Inherited from : py: class:AbstractIoULoss
__init__([factor_pos_class, reduction, ...])Init.
Inherited from : py: class:AbstractIoULike
__repr__()Return repr(self).
__str__()Return str(self).
Inherited from : py: class:Module
__init__([factor_pos_class, reduction, ...])Init.
__call__(*input, **kwargs)Call self as a function.
__setstate__(state)__getattr__(name)__setattr__(name, value)Implement setattr(self, name, value).
__delattr__(name)Implement delattr(self, name).
__repr__()Return repr(self).
__dir__()Default dir() implementation.
- Parameters
- __init__(factor_pos_class=0.5, reduction=BatchReduction.mean, target_thresh=0.0)[source]
Init.
- Parameters
target_thresh (float) – threshold to binarize targets
factor_pos_class (float) – balancing factor \(b\) in [0,1] applied to the foreground (1) class; defaults to 0.5 (i.e. equal weighting)
reduction (Union[BatchReduction, Callable[[Tensor], Tensor]]) – reduction method to aggregate the instance-wise results of the batch; must be a callable on a tensor which reduces the 0th dimension; for examples see instances of
BatchReduction
- factor_pos_class: float
Balancing factor \(b\) applied to the foreground (value 1) class. See loss formula (1).
- reduction: Union[BatchReduction, Callable[[Tensor], Tensor]]
Reduction method to aggregate the instance-wise results of the batch.