Table of Contents
Shortcuts

ignite.metrics#

Metrics provide a way to compute various quantities of interest in an online fashion without having to store the entire output history of a model.

In practice a user needs to attach the metric instance to an engine. The metric value is then computed using the output of the engine’s process_function:

from ignite.engine import Engine
from ignite.metrics import Accuracy

def process_function(engine, batch):
    # ...
    return y_pred, y

engine = Engine(process_function)
metric = Accuracy()
metric.attach(engine, "accuracy")
# ...
state = engine.run(data)
print(f"Accuracy: {state.metrics['accuracy']}")

If the engine’s output is not in the format (y_pred, y) or {'y_pred': y_pred, 'y': y, ...}, the user can use the output_transform argument to transform it:

from ignite.engine import Engine
from ignite.metrics import Accuracy

def process_function(engine, batch):
    # ...
    return {'y_pred': y_pred, 'y_true': y, ...}

engine = Engine(process_function)

def output_transform(output):
    # `output` variable is returned by above `process_function`
    y_pred = output['y_pred']
    y = output['y_true']
    return y_pred, y  # output format is according to `Accuracy` docs

metric = Accuracy(output_transform=output_transform)
metric.attach(engine, "accuracy")
# ...
state = engine.run(data)
print(f"Accuracy: {state.metrics['accuracy']}")

Warning

Please, be careful when using lambda functions to setup multiple output_transform for multiple metrics

# Wrong
# metrics_group = [Accuracy(output_transform=lambda output: output[name]) for name in names]
# As lambda can not store `name` and all `output_transform` will use the last `name`

# A correct way. For example, using functools.partial
from functools import partial

def ot_func(output, name):
    return output[name]

metrics_group = [Accuracy(output_transform=partial(ot_func, name=name)) for name in names]

For more details, see here

Note

Most of implemented metrics are adapted to distributed computations and reduce their internal states across supported devices before computing metric value. This can be helpful to run the evaluation on multiple nodes/GPU instances/TPUs with a distributed data sampler. Following code snippet shows in detail how to use metrics:

device = f"cuda:{local_rank}"
model = torch.nn.parallel.DistributedDataParallel(model,
                                                  device_ids=[local_rank, ],
                                                  output_device=local_rank)
test_sampler = DistributedSampler(test_dataset)
test_loader = DataLoader(
    test_dataset,
    batch_size=batch_size,
    sampler=test_sampler,
    num_workers=num_workers,
    pin_memory=True
)

evaluator = create_supervised_evaluator(model, metrics={'accuracy': Accuracy()}, device=device)

Note

Metrics cannot be serialized using pickle module because the implementation is based on lambda functions. Therefore, use the third party library dill to overcome the limitation of pickle.

Reset, Update, Compute API#

User can also call directly the following methods on the metric:

  • reset() : resets internal variables and accumulators

  • update() : updates internal variables and accumulators with provided batch output (y_pred, y)

  • compute() : computes custom metric and return the result

This API gives a more fine-grained/custom usage on how to compute a metric. For example:

from ignite.metrics import Precision

# Define the metric
precision = Precision()

# Start accumulation:
for x, y in data:
    y_pred = model(x)
    precision.update((y_pred, y))

# Compute the result
print("Precision: ", precision.compute())

# Reset metric
precision.reset()

# Start new accumulation:
for x, y in data:
    y_pred = model(x)
    precision.update((y_pred, y))

# Compute new result
print("Precision: ", precision.compute())

Metric arithmetics#

Metrics could be combined together to form new metrics. This could be done through arithmetics, such as metric1 + metric2, use PyTorch operators, such as (metric1 + metric2).pow(2).mean(), or use a lambda function, such as MetricsLambda(lambda a, b: torch.mean(a + b), metric1, metric2).

For example:

from ignite.metrics import Precision, Recall

precision = Precision(average=False)
recall = Recall(average=False)
F1 = (precision * recall * 2 / (precision + recall)).mean()

Note

This example computes the mean of F1 across classes. To combine precision and recall to get F1 or other F metrics, we have to be careful that average=False, i.e. to use the unaveraged precision and recall, otherwise we will not be computing F-beta metrics.

Metrics also support indexing operation (if metric’s result is a vector/matrix/tensor). For example, this can be useful to compute mean metric (e.g. precision, recall or IoU) ignoring the background:

from ignite.metrics import ConfusionMatrix

cm = ConfusionMatrix(num_classes=10)
iou_metric = IoU(cm)
iou_no_bg_metric = iou_metric[:9]  # We assume that the background index is 9
mean_iou_no_bg_metric = iou_no_bg_metric.mean()
# mean_iou_no_bg_metric.compute() -> tensor(0.12345)

How to create a custom metric#

To create a custom metric one needs to create a new class inheriting from Metric and override three methods :

  • reset() : resets internal variables and accumulators

  • update() : updates internal variables and accumulators with provided batch output (y_pred, y)

  • compute() : computes custom metric and return the result

For example, we would like to implement for illustration purposes a multi-class accuracy metric with some specific condition (e.g. ignore user-defined classes):

from ignite.metrics import Metric
from ignite.exceptions import NotComputableError

# These decorators helps with distributed settings
from ignite.metrics.metric import sync_all_reduce, reinit__is_reduced


class CustomAccuracy(Metric):

    def __init__(self, ignored_class, output_transform=lambda x: x, device="cpu"):
        self.ignored_class = ignored_class
        self._num_correct = None
        self._num_examples = None
        super(CustomAccuracy, self).__init__(output_transform=output_transform, device=device)

    @reinit__is_reduced
    def reset(self):
        self._num_correct = torch.tensor(0, device=self._device)
        self._num_examples = 0
        super(CustomAccuracy, self).reset()

    @reinit__is_reduced
    def update(self, output):
        y_pred, y = output[0].detach(), output[1].detach()

        indices = torch.argmax(y_pred, dim=1)

        mask = (y != self.ignored_class)
        mask &= (indices != self.ignored_class)
        y = y[mask]
        indices = indices[mask]
        correct = torch.eq(indices, y).view(-1)

        self._num_correct += torch.sum(correct).to(self._device)
        self._num_examples += correct.shape[0]

    @sync_all_reduce("_num_examples", "_num_correct")
    def compute(self):
        if self._num_examples == 0:
            raise NotComputableError('CustomAccuracy must have at least one example before it can be computed.')
        return self._num_correct.item() / self._num_examples

We imported necessary classes as Metric, NotComputableError and decorators to adapt the metric for distributed setting. In reset method, we reset internal variables _num_correct and _num_examples which are used to compute the custom metric. In updated method we define how to update the internal variables. And finally in compute method, we compute metric value.

Notice that _num_correct is a tensor, since in update we accumulate tensor values. _num_examples is a python scalar since we accumulate normal integers. For differentiable metrics, you must detach the accumulated values before adding them to the internal variables.

We can check this implementation in a simple case:

import torch
torch.manual_seed(8)

m = CustomAccuracy(ignored_class=3)

batch_size = 4
num_classes = 5

y_pred = torch.rand(batch_size, num_classes)
y = torch.randint(0, num_classes, size=(batch_size, ))

m.update((y_pred, y))
res = m.compute()

print(y, torch.argmax(y_pred, dim=1))
# Out: tensor([2, 2, 2, 3]) tensor([2, 1, 0, 0])

print(m._num_correct, m._num_examples, res)
# Out: 1 3 0.3333333333333333

Metrics and its usages#

By default, Metrics are epoch-wise, it means

  • reset() is triggered every EPOCH_STARTED (See Events).

  • update() is triggered every ITERATION_COMPLETED.

  • compute() is triggered every EPOCH_COMPLETED.

Usages can be user defined by creating a class inheriting for MetricUsage. See the list below of usages.

Complete list of usages#

Metrics and distributed computations#

In the above example, CustomAccuracy has reset, update, compute methods decorated with reinit__is_reduced(), sync_all_reduce(). The purpose of these features is to adapt metrics in distributed computations on supported backend and devices (see ignite.distributed for more details). More precisely, in the above example we added @sync_all_reduce("_num_examples", "_num_correct") over compute method. This means that when compute method is called, metric’s interal variables self._num_examples and self._num_correct are summed up over all participating devices. Therefore, once collected, these internal variables can be used to compute the final metric value.

Complete list of metrics#

Metric

Base class for all Metrics.

Accuracy

Calculates the accuracy for binary, multiclass and multilabel data.

Loss

Calculates the average loss according to the passed loss_fn.

MetricsLambda

Apply a function to other metrics to obtain a new metric.

MeanAbsoluteError

Calculates the mean absolute error.

MeanPairwiseDistance

Calculates the mean PairwiseDistance.

MeanSquaredError

Calculates the mean squared error.

ConfusionMatrix

Calculates confusion matrix for multi-class data.

TopKCategoricalAccuracy

Calculates the top-k categorical accuracy.

Average

Helper class to compute arithmetic average of a single variable.

DiceCoefficient

Calculates Dice Coefficient for a given ConfusionMatrix metric.

EpochMetric

Class for metrics that should be computed on the entire output history of a model.

Fbeta

Calculates F-beta score.

GeometricAverage

Helper class to compute geometric average of a single variable.

IoU

Calculates Intersection over Union using ConfusionMatrix metric.

mIoU

Calculates mean Intersection over Union using ConfusionMatrix metric.

Precision

Calculates precision for binary and multiclass data.

PSNR

Computes average Peak signal-to-noise ratio (PSNR).

Recall

Calculates recall for binary and multiclass data.

RootMeanSquaredError

Calculates the root mean squared error.

RunningAverage

Compute running average of a metric or the output of process function.

VariableAccumulation

Single variable accumulator helper to compute (arithmetic, geometric, harmonic) average of a single variable.

Frequency

Provides metrics for the number of examples processed per second.

SSIM

Computes Structual Similarity Index Measure
class ignite.metrics.Accuracy(output_transform=<function Accuracy.<lambda>>, is_multilabel=False, device=device(type='cpu'))[source]#

Calculates the accuracy for binary, multiclass and multilabel data.

Accuracy=TP+TNTP+TN+FP+FN\text{Accuracy} = \frac{ TP + TN }{ TP + TN + FP + FN }

where TP\text{TP} is true positives, TN\text{TN} is true negatives, FP\text{FP} is false positives and FN\text{FN} is false negatives.

  • update must receive output of the form (y_pred, y) or {'y_pred': y_pred, 'y': y}.

  • y_pred must be in the following shape (batch_size, num_categories, …) or (batch_size, …).

  • y must be in the following shape (batch_size, …).

  • y and y_pred must be in the following shape of (batch_size, num_categories, …) and num_categories must be greater than 1 for multilabel cases.

In binary and multilabel cases, the elements of y and y_pred should have 0 or 1 values. Thresholding of predictions can be done as below:

def thresholded_output_transform(output):
    y_pred, y = output
    y_pred = torch.round(y_pred)
    return y_pred, y

binary_accuracy = Accuracy(thresholded_output_transform)
Parameters

  • output_transform (callable, optional) – a callable that is used to transform the Engine’s process_function’s output into the form expected by the metric. This can be useful if, for example, you have a multi-output model and you want to compute the metric with respect to one of the outputs.

  • is_multilabel (bool, optional) – flag to use in multilabel case. By default, False.

  • device (str or torch.device) – specifies which device updates are accumulated on. Setting the metric’s device to be the same as your update arguments ensures the update method is non-blocking. By default, CPU.

class ignite.metrics.Average(output_transform=<function Average.<lambda>>, device=device(type='cpu'))[source]#

Helper class to compute arithmetic average of a single variable.

  • update must receive output of the form x.

  • x can be a number or torch.Tensor.

Note

Number of samples is updated following the rule:

  • +1 if input is a number

  • +1 if input is a 1D torch.Tensor

  • +batch_size if input is an ND torch.Tensor. Batch size is the first dimension (shape[0]).

For input x being an ND torch.Tensor with N > 1, the first dimension is seen as the number of samples and is summed up and added to the accumulator: accumulator += x.sum(dim=0)

Examples:

evaluator = ...

custom_var_mean = Average(output_transform=lambda output: output['custom_var'])
custom_var_mean.attach(evaluator, 'mean_custom_var')

state = evaluator.run(dataset)
# state.metrics['mean_custom_var'] -> average of output['custom_var']
Parameters

  • output_transform (callable, optional) – a callable that is used to transform the Engine’s process_function’s output into the form expected by the metric. This can be useful if, for example, you have a multi-output model and you want to compute the metric with respect to one of the outputs.

  • device (str or torch.device) – specifies which device updates are accumulated on. Setting the metric’s device to be the same as your update arguments ensures the update method is non-blocking. By default, CPU.

class ignite.metrics.ConfusionMatrix(num_classes, average=None, output_transform=<function ConfusionMatrix.<lambda>>, device=device(type='cpu'))[source]#

Calculates confusion matrix for multi-class data.

  • update must receive output of the form (y_pred, y) or {'y_pred': y_pred, 'y': y}.

  • y_pred must contain logits and has the following shape (batch_size, num_classes, …). If you are doing binary classification, see Note for an example on how to get this.

  • y should have the following shape (batch_size, …) and contains ground-truth class indices with or without the background class. During the computation, argmax of y_pred is taken to determine predicted classes.

Parameters

  • num_classes (int) – Number of classes, should be > 1. See notes for more details.

  • average (str, optional) – confusion matrix values averaging schema: None, “samples”, “recall”, “precision”. Default is None. If average=”samples” then confusion matrix values are normalized by the number of seen samples. If average=”recall” then confusion matrix values are normalized such that diagonal values represent class recalls. If average=”precision” then confusion matrix values are normalized such that diagonal values represent class precisions.

  • output_transform (callable, optional) – a callable that is used to transform the Engine’s process_function’s output into the form expected by the metric. This can be useful if, for example, you have a multi-output model and you want to compute the metric with respect to one of the outputs.

  • device (str or torch.device) – specifies which device updates are accumulated on. Setting the metric’s device to be the same as your update arguments ensures the update method is non-blocking. By default, CPU.

Note

In case of the targets y in (batch_size, …) format, target indices between 0 and num_classes only contribute to the confusion matrix and others are neglected. For example, if num_classes=20 and target index equal 255 is encountered, then it is filtered out.

If you are doing binary classification with a single output unit, you may have to transform your network output, so that you have one value for each class. E.g. you can transform your network output into a one-hot vector with:

def binary_one_hot_output_transform(output):
    y_pred, y = output
    y_pred = torch.sigmoid(y_pred).round().long()
    y_pred = ignite.utils.to_onehot(y_pred, 2)
    y = y.long()
    return y_pred, y

metrics = {
    "confusion_matrix": ConfusionMatrix(2, output_transform=binary_one_hot_output_transform),
}

evaluator = create_supervised_evaluator(
    model, metrics=metrics, output_transform=lambda x, y, y_pred: (y_pred, y)
)
ignite.metrics.DiceCoefficient(cm, ignore_index=None)[source]#

Calculates Dice Coefficient for a given ConfusionMatrix metric.

Parameters

  • cm (ConfusionMatrix) – instance of confusion matrix metric

  • ignore_index (int, optional) – index to ignore, e.g. background index

Return type

MetricsLambda

class ignite.metrics.EpochMetric(compute_fn, output_transform=<function EpochMetric.<lambda>>, check_compute_fn=True, device=device(type='cpu'))[source]#

Class for metrics that should be computed on the entire output history of a model. Model’s output and targets are restricted to be of shape (batch_size, n_classes). Output datatype should be float32. Target datatype should be long.

Warning

Current implementation stores all input data (output and target) in as tensors before computing a metric. This can potentially lead to a memory error if the input data is larger than available RAM.

In distributed configuration, all stored data (output and target) is mutually collected across all processes using all gather collective operation. This can potentially lead to a memory error. Compute method executes compute_fn on zero rank process only and final result is broadcasted to all processes.

  • update must receive output of the form (y_pred, y) or {'y_pred': y_pred, 'y': y}.

If target shape is (batch_size, n_classes) and n_classes > 1 than it should be binary: e.g. [[0, 1, 0, 1], ].

Parameters

  • compute_fn (callable) – a callable with the signature (torch.tensor, torch.tensor) takes as the input predictions and targets and returns a scalar. Input tensors will be on specified device (see arg below).

  • output_transform (callable, optional) – a callable that is used to transform the Engine’s process_function’s output into the form expected by the metric. This can be useful if, for example, you have a multi-output model and you want to compute the metric with respect to one of the outputs.

  • check_compute_fn (bool) – if True, compute_fn is run on the first batch of data to ensure there are no issues. If issues exist, user is warned that there might be an issue with the compute_fn. Default, True.

  • device (str or torch.device, optional) – optional device specification for internal storage.

Warning

EpochMetricWarning: User is warned that there are issues with compute_fn on a batch of data processed. To disable the warning, set check_compute_fn=False.

ignite.metrics.Fbeta(beta, average=True, precision=None, recall=None, output_transform=None, device=device(type='cpu'))[source]#

Calculates F-beta score.

Fβ=(1+β2)∗precision∗recall(β2∗precision)+recallF_\beta = \left( 1 + \beta^2 \right) * \frac{ \text{precision} * \text{recall} } { \left( \beta^2 * \text{precision} \right) + \text{recall} }

where β\beta is a positive real factor.

Parameters

  • beta (float) – weight of precision in harmonic mean

  • average (bool, optional) – if True, F-beta score is computed as the unweighted average (across all classes in multiclass case), otherwise, returns a tensor with F-beta score for each class in multiclass case.

  • precision (Precision, optional) – precision object metric with average=False to compute F-beta score

  • recall (Precision, optional) – recall object metric with average=False to compute F-beta score

  • output_transform (callable, optional) – a callable that is used to transform the Engine’s process_function’s output into the form expected by the metric. It is used only if precision or recall are not provided.

  • device (str or torch.device) – specifies which device updates are accumulated on. Setting the metric’s device to be the same as your update arguments ensures the update method is non-blocking. By default, CPU.

Returns

MetricsLambda, F-beta metric

Return type

MetricsLambda

class ignite.metrics.GeometricAverage(output_transform=<function GeometricAverage.<lambda>>, device=device(type='cpu'))[source]#

Helper class to compute geometric average of a single variable.

  • update must receive output of the form x.

  • x can be a positive number or a positive torch.Tensor, such that torch.log(x) is not nan.

Note

Number of samples is updated following the rule:

  • +1 if input is a number

  • +1 if input is a 1D torch.Tensor

  • +batch_size if input is a ND torch.Tensor. Batch size is the first dimension (shape[0]).

For input x being an ND torch.Tensor with N > 1, the first dimension is seen as the number of samples and is aggregated and added to the accumulator: accumulator *= prod(x, dim=0)

Parameters

  • output_transform (callable, optional) – a callable that is used to transform the Engine’s process_function’s output into the form expected by the metric. This can be useful if, for example, you have a multi-output model and you want to compute the metric with respect to one of the outputs.

  • device (str or torch.device) – specifies which device updates are accumulated on. Setting the metric’s device to be the same as your update arguments ensures the update method is non-blocking. By default, CPU.

ignite.metrics.IoU(cm, ignore_index=None)[source]#

Calculates Intersection over Union using ConfusionMatrix metric.

J(A,B)=∣A∩B∣∣A∪B∣\text{J}(A, B) = \frac{ \lvert A \cap B \rvert }{ \lvert A \cup B \rvert }
Parameters

  • cm (ConfusionMatrix) – instance of confusion matrix metric

  • ignore_index (int, optional) – index to ignore, e.g. background index

Returns

MetricsLambda

Return type

MetricsLambda

Examples:

train_evaluator = ...

cm = ConfusionMatrix(num_classes=num_classes)
IoU(cm, ignore_index=0).attach(train_evaluator, 'IoU')

state = train_evaluator.run(train_dataset)
# state.metrics['IoU'] -> tensor of shape (num_classes - 1, )
ignite.metrics.mIoU(cm, ignore_index=None)[source]#

Calculates mean Intersection over Union using ConfusionMatrix metric.

Parameters

  • cm (ConfusionMatrix) – instance of confusion matrix metric

  • ignore_index (int, optional) – index to ignore, e.g. background index

Returns

MetricsLambda

Return type

MetricsLambda

Examples:

train_evaluator = ...

cm = ConfusionMatrix(num_classes=num_classes)
mIoU(cm, ignore_index=0).attach(train_evaluator, 'mean IoU')

state = train_evaluator.run(train_dataset)
# state.metrics['mean IoU'] -> scalar
class ignite.metrics.Loss(loss_fn, output_transform=<function Loss.<lambda>>, batch_size=<function Loss.<lambda>>, device=device(type='cpu'))[source]#

Calculates the average loss according to the passed loss_fn.

Parameters

  • loss_fn (callable) – a callable taking a prediction tensor, a target tensor, optionally other arguments, and returns the average loss over all observations in the batch.

  • output_transform (callable) – a callable that is used to transform the Engine’s process_function’s output into the form expected by the metric. This can be useful if, for example, you have a multi-output model and you want to compute the metric with respect to one of the outputs. The output is expected to be a tuple (prediction, target) or (prediction, target, kwargs) where kwargs is a dictionary of extra keywords arguments. If extra keywords arguments are provided they are passed to loss_fn.

  • batch_size (callable) – a callable taking a target tensor that returns the first dimension size (usually the batch size).

  • device (str or torch.device) – specifies which device updates are accumulated on. Setting the metric’s device to be the same as your update arguments ensures the update method is non-blocking. By default, CPU.

class ignite.metrics.MeanAbsoluteError(output_transform=<function Metric.<lambda>>, device=device(type='cpu'))[source]#

Calculates the mean absolute error.

MAE=1N∑i=1N∣yi−xi∣\text{MAE} = \frac{1}{N} \sum_{i=1}^N \lvert y_{i} - x_{i} \rvert

where yiy_{i} is the prediction tensor and xix_{i} is ground true tensor.

  • update must receive output of the form (y_pred, y) or {'y_pred': y_pred, 'y': y}.

Parameters
class ignite.metrics.MeanPairwiseDistance(p=2, eps=1e-06, output_transform=<function MeanPairwiseDistance.<lambda>>, device=device(type='cpu'))[source]#

Calculates the mean PairwiseDistance. Average of pairwise distances computed on provided batches.

  • update must receive output of the form (y_pred, y) or {'y_pred': y_pred, 'y': y}.

Parameters
class ignite.metrics.MeanSquaredError(output_transform=<function Metric.<lambda>>, device=device(type='cpu'))[source]#

Calculates the mean squared error.

MSE=1N∑i=1N(yi−xi)2\text{MSE} = \frac{1}{N} \sum_{i=1}^N \left(y_{i} - x_{i} \right)^2

where yiy_{i} is the prediction tensor and xix_{i} is ground true tensor.

  • update must receive output of the form (y_pred, y) or {'y_pred': y_pred, 'y': y}.

Parameters
class ignite.metrics.Metric(output_transform=<function Metric.<lambda>>, device=device(type='cpu'))[source]#

Base class for all Metrics.

Parameters

  • output_transform (callable, optional) – a callable that is used to transform the Engine’s process_function’s output into the form expected by the metric. This can be useful if, for example, you have a multi-output model and you want to compute the metric with respect to one of the outputs. By default, metrics require the output as (y_pred, y) or {'y_pred': y_pred, 'y': y}.

  • device (str or torch.device) – specifies which device updates are accumulated on. Setting the metric’s device to be the same as your update arguments ensures the update method is non-blocking. By default, CPU.

required_output_keys#

dictionary defines required keys to be found in engine.state.output if the latter is a dictionary. Default, ("y_pred", "y"). This is useful with custom metrics that can require other arguments than predictions y_pred and targets y. See notes below for an example.

Type

tuple

Note

Let’s implement a custom metric that requires y_pred, y and x as input for update function. In the example below we show how to setup standard metric like Accuracy and the custom metric using by an evaluator created with create_supervised_evaluator() method.

# https://discuss.pytorch.org/t/how-access-inputs-in-custom-ignite-metric/91221/5

import torch
import torch.nn as nn

from ignite.metrics import Metric, Accuracy
from ignite.engine import create_supervised_evaluator

class CustomMetric(Metric):

    required_output_keys = ("y_pred", "y", "x")

    def __init__(self, *args, **kwargs):
        super().__init__(*args, **kwargs)

    def update(self, output):
        y_pred, y, x = output
        # ...

    def reset(self):
        # ...
        pass

    def compute(self):
        # ...
        pass

model = ...

metrics = {
    "Accuracy": Accuracy(),
    "CustomMetric": CustomMetric()
}

evaluator = create_supervised_evaluator(
    model,
    metrics=metrics,
    output_transform=lambda x, y, y_pred: {"x": x, "y": y, "y_pred": y_pred}
)

res = evaluator.run(data)

Changed in version 0.4.2: required_output_keys became public attribute.

attach(engine, name, usage=<ignite.metrics.metric.EpochWise object>)[source]#

Attaches current metric to provided engine. On the end of engine’s run, engine.state.metrics dictionary will contain computed metric’s value under provided name.

Parameters
Return type

None

Example:

metric = ...
metric.attach(engine, "mymetric")

assert "mymetric" in engine.run(data).metrics

assert metric.is_attached(engine)

Example with usage:

metric = ...
metric.attach(engine, "mymetric", usage=BatchWise.usage_name)

assert "mymetric" in engine.run(data).metrics

assert metric.is_attached(engine, usage=BatchWise.usage_name)
completed(engine, name)[source]#

Helper method to compute metric’s value and put into the engine. It is automatically attached to the engine with attach().

Parameters

  • engine (Engine) – the engine to which the metric must be attached

  • name (str) – the name of the metric used as key in dict engine.state.metrics

Return type

None

Changed in version 0.4.3: Added dict in metrics results.

abstract compute()[source]#

Computes the metric based on it’s accumulated state.

By default, this is called at the end of each epoch.

Returns

the actual quantity of interest. However, if a Mapping is returned, it will be (shallow) flattened into engine.state.metrics when completed() is called.

Return type

Any

Raises

NotComputableError – raised when the metric cannot be computed.

detach(engine, usage=<ignite.metrics.metric.EpochWise object>)[source]#

Detaches current metric from the engine and no metric’s computation is done during the run. This method in conjunction with attach() can be useful if several metrics need to be computed with different periods. For example, one metric is computed every training epoch and another metric (e.g. more expensive one) is done every n-th training epoch.

Parameters

  • engine (Engine) – the engine from which the metric must be detached

  • usage (str or MetricUsage, optional) – the usage of the metric. Valid string values should be ‘epoch_wise’ (default) or ‘batch_wise’.

Return type

None

Example:

metric = ...
engine = ...
metric.detach(engine)

assert "mymetric" not in engine.run(data).metrics

assert not metric.is_attached(engine)

Example with usage:

metric = ...
engine = ...
metric.detach(engine, usage="batch_wise")

assert "mymetric" not in engine.run(data).metrics

assert not metric.is_attached(engine, usage="batch_wise")
is_attached(engine, usage=<ignite.metrics.metric.EpochWise object>)[source]#

Checks if current metric is attached to provided engine. If attached, metric’s computed value is written to engine.state.metrics dictionary.

Parameters

  • engine (Engine) – the engine checked from which the metric should be attached

  • usage (str or MetricUsage, optional) – the usage of the metric. Valid string values should be ‘epoch_wise’ (default) or ‘batch_wise’.

Return type

bool

iteration_completed(engine)[source]#

Helper method to update metric’s computation. It is automatically attached to the engine with attach().

Parameters

engine (Engine) – the engine to which the metric must be attached

Return type

None

abstract reset()[source]#

Resets the metric to it’s initial state.

By default, this is called at the start of each epoch.

Return type

None

started(engine)[source]#

Helper method to start data gathering for metric’s computation. It is automatically attached to the engine with attach().

Parameters

engine (Engine) – the engine to which the metric must be attached

Return type

None

abstract update(output)[source]#

Updates the metric’s state using the passed batch output.

By default, this is called once for each batch.

Parameters

output (Any) – the is the output from the engine’s process function.

Return type

None

class ignite.metrics.MetricsLambda(f, *args, **kwargs)[source]#

Apply a function to other metrics to obtain a new metric. The result of the new metric is defined to be the result of applying the function to the result of argument metrics.

When update, this metric does not recursively update the metrics it depends on. When reset, all its dependency metrics would be resetted. When attach, all its dependency metrics would be attached automatically (but partially, e.g is_attached() will return False).

Parameters

  • f (callable) – the function that defines the computation

  • args (sequence) – Sequence of other metrics or something else that will be fed to f as arguments.

  • kwargs (Any) –

Example:

precision = Precision(average=False)
recall = Recall(average=False)

def Fbeta(r, p, beta):
    return torch.mean((1 + beta ** 2) * p * r / (beta ** 2 * p + r + 1e-20)).item()

F1 = MetricsLambda(Fbeta, recall, precision, 1)
F2 = MetricsLambda(Fbeta, recall, precision, 2)
F3 = MetricsLambda(Fbeta, recall, precision, 3)
F4 = MetricsLambda(Fbeta, recall, precision, 4)

When check if the metric is attached, if one of its dependency metrics is detached, the metric is considered detached too.

engine = ...
precision = Precision(average=False)

aP = precision.mean()

aP.attach(engine, "aP")

assert aP.is_attached(engine)
# partially attached
assert not precision.is_attached(engine)

precision.detach(engine)

assert not aP.is_attached(engine)
# fully attached
assert not precision.is_attached(engine)
class ignite.metrics.PSNR(data_range, output_transform=<function PSNR.<lambda>>, device=device(type='cpu'))[source]#

Computes average Peak signal-to-noise ratio (PSNR).

PSNR(I,J)=10∗log⁡10(MAXI2 MSE )\text{PSNR}(I, J) = 10 * \log_{10}\left(\frac{ MAX_{I}^2 }{ \text{ MSE } }\right)

where MSE\text{MSE} is mean squared error.

  • y_pred and y must have (batch_size, …) shape.

  • y_pred and y must have same dtype and same shape.

Parameters

  • data_range (int or float) – The data range of the target image (distance between minimum and maximum possible values). For other data types, please set the data range, otherwise an exception will be raised.

  • output_transform (callable, optional) – A callable that is used to transform the Engine’s process_function’s output into the form expected by the metric.

  • device (str or torch.device) – specifies which device updates are accumulated on. Setting the metric’s device to be the same as your update arguments ensures the update method is non-blocking. By default, CPU.

Example:

To use with Engine and process_function, simply attach the metric instance to the engine. The output of the engine’s process_function needs to be in format of (y_pred, y) or {'y_pred': y_pred, 'y': y, ...}.

def process_function(engine, batch):
    # ...
    return y_pred, y
engine = Engine(process_function)
psnr = PSNR(data_range=1.0)
psnr.attach(engine, "psnr")
# ...
state = engine.run(data)
print(f"PSNR: {state.metrics['psnr']}")

This metric by default accepts Grayscale or RGB images. But if you have YCbCr or YUV images, only Y channel is needed for computing PSNR. And, this can be done with output_transform. For instance,

def get_y_channel(output):
    y_pred, y = output
    # y_pred and y are (B, 3, H, W) and YCbCr or YUV images
    # let's select y channel
    return y_pred[:, 0, ...], y[:, 0, ...]

psnr = PSNR(data_range=219, output_transform=get_y_channel)
psnr.attach(engine, "psnr")
# ...
state = engine.run(data)
print(f"PSNR: {state.metrics['psrn']}")

New in version 0.4.3.

class ignite.metrics.Precision(output_transform=<function Precision.<lambda>>, average=False, is_multilabel=False, device=device(type='cpu'))[source]#

Calculates precision for binary and multiclass data.

Precision=TPTP+FP\text{Precision} = \frac{ TP }{ TP + FP }

where TP\text{TP} is true positives and FP\text{FP} is false positives.

  • update must receive output of the form (y_pred, y) or {'y_pred': y_pred, 'y': y}.

  • y_pred must be in the following shape (batch_size, num_categories, …) or (batch_size, …).

  • y must be in the following shape (batch_size, …).

In binary and multilabel cases, the elements of y and y_pred should have 0 or 1 values. Thresholding of predictions can be done as below:

def thresholded_output_transform(output):
    y_pred, y = output
    y_pred = torch.round(y_pred)
    return y_pred, y

precision = Precision(output_transform=thresholded_output_transform)

In multilabel cases, average parameter should be True. However, if user would like to compute F1 metric, for example, average parameter should be False. This can be done as shown below:

precision = Precision(average=False, is_multilabel=True)
recall = Recall(average=False, is_multilabel=True)
F1 = precision * recall * 2 / (precision + recall + 1e-20)
F1 = MetricsLambda(lambda t: torch.mean(t).item(), F1)

Warning

In multilabel cases, if average is False, current implementation stores all input data (output and target) in as tensors before computing a metric. This can potentially lead to a memory error if the input data is larger than available RAM.

Warning

In multilabel cases, if average is False, current implementation does not work with distributed computations. Results are not reduced across the GPUs. Computed result corresponds to the local rank’s (single GPU) result.

Parameters

  • output_transform (callable, optional) – a callable that is used to transform the Engine’s process_function’s output into the form expected by the metric. This can be useful if, for example, you have a multi-output model and you want to compute the metric with respect to one of the outputs.

  • average (bool, optional) – if True, precision is computed as the unweighted average (across all classes in multiclass case), otherwise, returns a tensor with the precision (for each class in multiclass case).

  • is_multilabel (bool, optional) – parameter should be True and the average is computed across samples, instead of classes.

  • device (str or torch.device) – specifies which device updates are accumulated on. Setting the metric’s device to be the same as your update arguments ensures the update method is non-blocking. By default, CPU.

class ignite.metrics.Recall(output_transform=<function Recall.<lambda>>, average=False, is_multilabel=False, device=device(type='cpu'))[source]#

Calculates recall for binary and multiclass data.

Recall=TPTP+FN\text{Recall} = \frac{ TP }{ TP + FN }

where TP\text{TP} is true positives and FN\text{FN} is false negatives.

  • update must receive output of the form (y_pred, y) or {'y_pred': y_pred, 'y': y}.

  • y_pred must be in the following shape (batch_size, num_categories, …) or (batch_size, …).

  • y must be in the following shape (batch_size, …).

In binary and multilabel cases, the elements of y and y_pred should have 0 or 1 values. Thresholding of predictions can be done as below:

def thresholded_output_transform(output):
    y_pred, y = output
    y_pred = torch.round(y_pred)
    return y_pred, y

recall = Recall(output_transform=thresholded_output_transform)

In multilabel cases, average parameter should be True. However, if user would like to compute F1 metric, for example, average parameter should be False. This can be done as shown below:

precision = Precision(average=False, is_multilabel=True)
recall = Recall(average=False, is_multilabel=True)
F1 = precision * recall * 2 / (precision + recall + 1e-20)
F1 = MetricsLambda(lambda t: torch.mean(t).item(), F1)

Warning

In multilabel cases, if average is False, current implementation stores all input data (output and target) in as tensors before computing a metric. This can potentially lead to a memory error if the input data is larger than available RAM.

Warning

In multilabel cases, if average is False, current implementation does not work with distributed computations. Results are not reduced across the GPUs. Computed result corresponds to the local rank’s (single GPU) result.

Parameters

  • output_transform (callable, optional) – a callable that is used to transform the Engine’s process_function’s output into the form expected by the metric. This can be useful if, for example, you have a multi-output model and you want to compute the metric with respect to one of the outputs.

  • average (bool, optional) – if True, precision is computed as the unweighted average (across all classes in multiclass case), otherwise, returns a tensor with the precision (for each class in multiclass case).

  • is_multilabel (bool, optional) – parameter should be True and the average is computed across samples, instead of classes.

  • device (str or torch.device) – specifies which device updates are accumulated on. Setting the metric’s device to be the same as your update arguments ensures the update method is non-blocking. By default, CPU.

class ignite.metrics.RootMeanSquaredError(output_transform=<function Metric.<lambda>>, device=device(type='cpu'))[source]#

Calculates the root mean squared error.

RMSE=1N∑i=1N(yi−xi)2\text{RMSE} = \sqrt{ \frac{1}{N} \sum_{i=1}^N \left(y_{i} - x_{i} \right)^2 }

where yiy_{i} is the prediction tensor and xix_{i} is ground true tensor.

  • update must receive output of the form (y_pred, y) or {‘y_pred’: y_pred, ‘y’: y}.

Parameters
class ignite.metrics.RunningAverage(src=None, alpha=0.98, output_transform=None, epoch_bound=True, device=None)[source]#

Compute running average of a metric or the output of process function.

Parameters

  • src (Metric or None) – input source: an instance of Metric or None. The latter corresponds to engine.state.output which holds the output of process function.

  • alpha (float, optional) – running average decay factor, default 0.98

  • output_transform (callable, optional) – a function to use to transform the output if src is None and corresponds the output of process function. Otherwise it should be None.

  • epoch_bound (boolean, optional) – whether the running average should be reset after each epoch (defaults to True).

  • device (str or torch.device, optional) – specifies which device updates are accumulated on. Should be None when src is an instance of Metric, as the running average will use the src’s device. Otherwise, defaults to CPU. Only applicable when the computed value from the metric is a tensor.

Examples:

alpha = 0.98
acc_metric = RunningAverage(Accuracy(output_transform=lambda x: [x[1], x[2]]), alpha=alpha)
acc_metric.attach(trainer, 'running_avg_accuracy')

avg_output = RunningAverage(output_transform=lambda x: x[0], alpha=alpha)
avg_output.attach(trainer, 'running_avg_loss')

@trainer.on(Events.ITERATION_COMPLETED)
def log_running_avg_metrics(engine):
    print("running avg accuracy:", engine.state.metrics['running_avg_accuracy'])
    print("running avg loss:", engine.state.metrics['running_avg_loss'])
class ignite.metrics.SSIM(data_range, kernel_size=(11, 11), sigma=(1.5, 1.5), k1=0.01, k2=0.03, gaussian=True, output_transform=<function SSIM.<lambda>>, device=device(type='cpu'))[source]#

Computes Structual Similarity Index Measure

Parameters

  • data_range (int or float) – Range of the image. Typically, 1.0 or 255.

  • kernel_size (int or list or tuple of int) – Size of the kernel. Default: (11, 11)

  • sigma (float or list or tuple of float) – Standard deviation of the gaussian kernel. Argument is used if gaussian=True. Default: (1.5, 1.5)

  • k1 (float) – Parameter of SSIM. Default: 0.01

  • k2 (float) – Parameter of SSIM. Default: 0.03

  • gaussian (bool) – True to use gaussian kernel, False to use uniform kernel

  • output_transform (callable, optional) – A callable that is used to transform the Engine’s process_function’s output into the form expected by the metric.

  • device (str or torch.device) – specifies which device updates are accumulated on. Setting the metric’s device to be the same as your update arguments ensures the update method is non-blocking. By default, CPU.

Example:

To use with Engine and process_function, simply attach the metric instance to the engine. The output of the engine’s process_function needs to be in the format of (y_pred, y) or {'y_pred': y_pred, 'y': y, ...}.

y_pred and y can be un-normalized or normalized image tensors. Depending on that, the user might need to adjust data_range. y_pred and y should have the same shape.

def process_function(engine, batch):
    # ...
    return y_pred, y
engine = Engine(process_function)
metric = SSIM(data_range=1.0)
metric.attach(engine, "ssim")

New in version 0.4.2.

class ignite.metrics.TopKCategoricalAccuracy(k=5, output_transform=<function TopKCategoricalAccuracy.<lambda>>, device=device(type='cpu'))[source]#

Calculates the top-k categorical accuracy.

  • update must receive output of the form (y_pred, y) or {'y_pred': y_pred, 'y': y}.

Parameters
class ignite.metrics.VariableAccumulation(op, output_transform=<function VariableAccumulation.<lambda>>, device=device(type='cpu'))[source]#

Single variable accumulator helper to compute (arithmetic, geometric, harmonic) average of a single variable.

  • update must receive output of the form x.

  • x can be a number or torch.Tensor.

Note

The class stores input into two public variables: accumulator and num_examples. Number of samples is updated following the rule:

  • +1 if input is a number

  • +1 if input is a 1D torch.Tensor

  • +batch_size if input is a ND torch.Tensor. Batch size is the first dimension (shape[0]).

Parameters

  • op (callable) – a callable to update accumulator. Method’s signature is (accumulator, output). For example, to compute arithmetic mean value, op = lambda a, x: a + x.

  • output_transform (callable, optional) – a callable that is used to transform the Engine’s process_function’s output into the form expected by the metric. This can be useful if, for example, you have a multi-output model and you want to compute the metric with respect to one of the outputs.

  • device (str or torch.device) – specifies which device updates are accumulated on. Setting the metric’s device to be the same as your update arguments ensures the update method is non-blocking. By default, CPU.

class ignite.metrics.MetricUsage(started, completed, iteration_completed)[source]#

Base class for all usages of metrics.

A usage of metric defines the events when a metric starts to compute, updates and completes. Valid events are from Events.

Parameters
class ignite.metrics.EpochWise[source]#

Epoch-wise usage of Metrics. It’s the default and most common usage of metrics.

Metric’s methods are triggered on the following engine events:

usage_name#

usage name string

Type

str

class ignite.metrics.BatchWise[source]#

Batch-wise usage of Metrics.

Metric’s methods are triggered on the following engine events:

usage_name#

usage name string

Type

str

class ignite.metrics.BatchFiltered(*args, **kwargs)[source]#

Batch filtered usage of Metrics. This usage is similar to epoch-wise but update event is filtered.

Metric’s methods are triggered on the following engine events:

Parameters

  • *args (Any) – Positional arguments to setup ITERATION_COMPLETED(*args, **kwargs)

  • **kwargs (Any) – Keyword arguments to setup ITERATION_COMPLETED(*args, **kwargs) handled by iteration_completed().

ignite.metrics.metric.reinit__is_reduced(func)[source]#

Helper decorator for distributed configuration.

See ignite.metrics on how to use it.

Parameters

func (Callable) –

Return type

Callable

ignite.metrics.metric.sync_all_reduce(*attrs)[source]#

Helper decorator for distributed configuration to collect instance attribute value across all participating processes.

See ignite.metrics on how to use it.

Parameters

*attrs (Any) – attribute names of decorated class

Return type

Callable