nntoolbox.vision.losses.metrics module

class nntoolbox.vision.losses.metrics.AngularLoss(alpha=45)

Bases: torch.nn.modules.module.Module

Based on https://github.com/leeesangwon/PyTorch-Image-Retrieval/blob/public/losses.py

forward(data: Tuple[torch.Tensor, torch.Tensor]) → torch.Tensor

Defines 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.

training: bool

class nntoolbox.vision.losses.metrics.ContrastiveLoss(margin=1.0)

Bases: torch.nn.modules.module.Module

Contrastive loss function.

Based on:

https://github.com/delijati/pytorch-siamese/blob/master/contrastive.py#L20

check_type_forward(in_types)

dist(x_0, x_1, eps=1e-08, squared=False)

forward(data: Tuple[torch.Tensor, …]) → torch.Tensor

Defines 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.

training: bool

class nntoolbox.vision.losses.metrics.NPairAngular(alpha=45, reg_lambda=0.002, angular_lambda=2)

Bases: torch.nn.modules.module.Module

Combining N-Pair loss and Angular loss

forward(data: Tuple[torch.Tensor, torch.Tensor]) → torch.Tensor

Defines 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.

training: bool

class nntoolbox.vision.losses.metrics.NPairLoss(reg_lambda: float = 0.002)

Bases: torch.nn.modules.module.Module

forward(data: Tuple[torch.Tensor, torch.Tensor]) → torch.Tensor

Defines 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.

training: bool

class nntoolbox.vision.losses.metrics.TripletMarginLossV2(margin=1.0, p=2.0, eps=1e-06, swap=False, size_average=None, reduce=None, reduction='mean')

Bases: torch.nn.modules.loss.TripletMarginLoss

A quick wrapper for margin loss

eps: float

forward(data: Tuple[torch.Tensor, torch.Tensor, torch.Tensor]) → torch.Tensor

Defines 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.

margin: float

p: float

swap: bool

class nntoolbox.vision.losses.metrics.TripletSoftMarginLoss(p: float = 2.0)

Bases: torch.nn.modules.module.Module

forward(data: Tuple[torch.Tensor, torch.Tensor, torch.Tensor]) → torch.Tensor

Defines 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.

training: bool

class nntoolbox.vision.losses.metrics.VerificationLoss(embedding_dim: int)

Bases: torch.nn.modules.module.Module

Verify if two embeddings belong to the same class

forward(data: Tuple[torch.Tensor, …]) → torch.Tensor

Defines 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.

get_verifier()

training: bool