Frechet Inception Distance (FID)¶
Module Interface¶
- class torchmetrics.image.fid.FrechetInceptionDistance(feature=2048, reset_real_features=True, compute_on_step=None, **kwargs)[source]
Calculates Fréchet inception distance (FID) which is used to access the quality of generated images. Given by
where is the multivariate normal distribution estimated from Inception v3 [1] features calculated on real life images and is the multivariate normal distribution estimated from Inception v3 features calculated on generated (fake) images. 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. The boolian flagreal
determines if the images should update the statistics of the real distribution or the fake distribution.Note
using this metrics requires you to have
scipy
install. Either install aspip install torchmetrics[image]
orpip install scipy
Note
using this metric with the default feature extractor requires that
torch-fidelity
is installed. Either install aspip install torchmetrics[image]
orpip install torch-fidelity
Note
the
forward
method can be used butcompute_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 defaultforward
will just callupdate
underneat.- Parameters
feature¶ (
Union
[int
,Module
]) –Either an integer or
nn.Module
:an integer will indicate the inceptionv3 feature layer to choose. Can be one of the following: 64, 192, 768, 2048
an
nn.Module
for using a custom feature extractor. Expects that its forward method returns an[N,d]
matrix whereN
is the batch size andd
is the feature size.
reset_real_features¶ (
bool
) – Whether to also reset the real features. Since in many cases the real dataset does not change, the features can cached them to avoid recomputing them which is costly. Set this toFalse
if your dataset does not change.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] Rethinking the Inception Architecture for Computer Vision Christian Szegedy, Vincent Vanhoucke, Sergey Ioffe, Jonathon Shlens, Zbigniew Wojna https://arxiv.org/abs/1512.00567
[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 anint
(default settings) andtorch-fidelity
is not installedValueError – If
feature
is set to anint
not in [64, 192, 768, 2048]TypeError – If
feature
is not anstr
,int
ortorch.nn.Module
ValueError – If
reset_real_features
is not anbool
Example
>>> import torch >>> _ = torch.manual_seed(123) >>> from torchmetrics.image.fid import FrechetInceptionDistance >>> fid = FrechetInceptionDistance(feature=64) >>> # generate two slightly overlapping image intensity distributions >>> imgs_dist1 = torch.randint(0, 200, (100, 3, 299, 299), dtype=torch.uint8) >>> imgs_dist2 = torch.randint(100, 255, (100, 3, 299, 299), dtype=torch.uint8) >>> fid.update(imgs_dist1, real=True) >>> fid.update(imgs_dist2, real=False) >>> fid.compute() tensor(12.7202)
Initializes internal Module state, shared by both nn.Module and ScriptModule.
- compute()[source]
Calculate FID score based on accumulated extracted features from the two distributions.
- Return type
- reset()[source]
This method automatically resets the metric state variables to their default value.
- Return type