Inception Score

Module Interface

class torchmetrics.image.inception.InceptionScore(feature='logits_unbiased', splits=10, compute_on_step=None, **kwargs)

Calculates the Inception Score (IS) which is used to access how realistic generated images are. It is defined as

where is the KL divergence between the conditional distribution and the margianl distribution . Both the conditional and marginal distribution is calculated from features extracted from the images. The score is calculated on random splits of the images such that both a mean and standard deviation of the score are returned. The metric was originally proposed in [1].

Using the default feature extraction (Inception v3 using the original weights from [2]), the input is expected to be mini-batches of 3-channel RGB images of shape (3 x H x W) with dtype uint8. All images will be resized to 299 x 299 which is the size of the original training data.

Note

using this metric with the default feature extractor requires that torch-fidelity is installed. Either install as pip install torchmetrics[image] or pip install torch-fidelity

Note

the forward method can be used but compute_on_step is disabled by default (oppesit of all other metrics) as this metric does not really make sense to calculate on a single batch. This means that by default forward will just call update underneat.

Parameters

• feature (Union[str, int, Module]) –

  Either an str, integer or nn.Module:

  • an str or integer will indicate the inceptionv3 feature layer to choose. Can be one of the following: ‘logits_unbiased’, 64, 192, 768, 2048

  • an nn.Module for using a custom feature extractor. Expects that its forward method returns an [N,d] matrix where N is the batch size and d is the feature size.

• splits (int) – integer determining how many splits the inception score calculation should be split among

• compute_on_step (Optional[bool]) –

  Forward only calls update() and returns None if this is set to False.

  Deprecated since version v0.8: Argument has no use anymore and will be removed v0.9.

• kwargs (Dict[str, Any]) – Additional keyword arguments, see Advanced metric settings for more info.

References

[1] Improved Techniques for Training GANs Tim Salimans, Ian Goodfellow, Wojciech Zaremba, Vicki Cheung, Alec Radford, Xi Chen https://arxiv.org/abs/1606.03498

[2] GANs Trained by a Two Time-Scale Update Rule Converge to a Local Nash Equilibrium, Martin Heusel, Hubert Ramsauer, Thomas Unterthiner, Bernhard Nessler, Sepp Hochreiter https://arxiv.org/abs/1706.08500

Raises

• ValueError – If feature is set to an str or int and torch-fidelity is not installed

• ValueError – If feature is set to an str or int and not one of ['logits_unbiased', 64, 192, 768, 2048]

• TypeError – If feature is not an str, int or torch.nn.Module

Example

>>> import torch
>>> _ = torch.manual_seed(123)
>>> from torchmetrics.image.inception import InceptionScore
>>> inception = InceptionScore()
>>> # generate some images
>>> imgs = torch.randint(0, 255, (100, 3, 299, 299), dtype=torch.uint8)
>>> inception.update(imgs)
>>> inception.compute()
(tensor(1.0544), tensor(0.0117))

Initializes internal Module state, shared by both nn.Module and ScriptModule.

compute()

Override this method to compute the final metric value from state variables synchronized across the distributed backend.

Return type

Tuple[Tensor, Tensor]

update(imgs)

Update the state with extracted features.

Parameters

imgs (Tensor) – tensor with images feed to the feature extractor

Return type

None