i$ddlmZddlmZmZmZddlZddlmZddlm Z ddl m Z ddl m Z ddlmZdd lmZdd lmZmZdd lmZmZesd gZd dgiZGdde Zy))Sequence)AnyOptionalUnionN)Tensor)Module)NoTrainInceptionV3)Metric)rank_zero_warn) dim_zero_cat)_MATPLOTLIB_AVAILABLE_TORCH_FIDELITY_AVAILABLE)_AX_TYPE_PLOT_OUT_TYPEInceptionScore.plot)InceptionScorertorch_fidelityc eZdZUdZdZeed<dZeed<dZeed<dZ e ed<e ed <e ed <d Z eed < dd eeee fd edededdf fd ZdeddfdZdeeeffdZ ddeeeeefdeedefdZxZS)ra Calculate the Inception Score (IS) which is used to access how realistic generated images are. .. math:: IS = exp(\mathbb{E}_x KL(p(y | x ) || p(y))) where :math:`KL(p(y | x) || p(y))` is the KL divergence between the conditional distribution :math:`p(y|x)` and the marginal distribution :math:`p(y)`. 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 `inception ref1`_. Using the default feature extraction (Inception v3 using the original weights from `inception ref2`_), the input is expected to be mini-batches of 3-channel RGB images of shape ``(3xHxW)``. If argument ``normalize`` is ``True`` images are expected to be dtype ``float`` and have values in the ``[0,1]`` range, else if ``normalize`` is set to ``False`` images are expected to have dtype uint8 and take values in the ``[0, 255]`` range. All images will be resized to 299 x 299 which is the size of the original training data. .. hint:: 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`` As input to ``forward`` and ``update`` the metric accepts the following input - ``imgs`` (:class:`~torch.Tensor`): tensor with images feed to the feature extractor As output of `forward` and `compute` the metric returns the following output - ``inception_mean`` (:class:`~torch.Tensor`): float scalar tensor with mean inception score over subsets - ``inception_std`` (:class:`~torch.Tensor`): float scalar tensor with standard deviation of inception score over subsets Args: feature: 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: integer determining how many splits the inception score calculation should be split among kwargs: Additional keyword arguments, see :ref:`Metric kwargs` for more info. 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: >>> from torch import rand >>> 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(1.0549), tensor(0.0121)) Fis_differentiableThigher_is_betterfull_state_updategplot_lower_boundfeatures inceptionfeature_networkfeaturesplits normalizekwargsreturnNc t|d i|tdtt |t t frFts tdd}||vrtd|d|dtdt |g|_ n#t |tr||_ n td t |ts td ||_||_|j#d gd y)NzMetric `InceptionScore` will save all extracted features in buffer. For large datasets this may lead to large memory footprint.zInceptionScore metric requires that `Torch-fidelity` is installed. Either install as `pip install torchmetrics[image]` or `pip install torch-fidelity`.)logits_unbiased@iiz3Integer input to argument `feature` must be one of z , but got .zinception-v3-compat)name features_listz'Got unknown input to argument `feature`z*Argument `normalize` expected to be a boolr)dist_reduce_fx)super__init__r UserWarning isinstancestrintrModuleNotFoundError ValueErrorr rr TypeErrorboolrr add_state)selfrrrrvalid_int_input __class__s q/Volumes/fast/ai/experiments/voice-extract-mac/.venv/lib/python3.12/site-packages/torchmetrics/image/inception.pyr+zInceptionScore.__init__ms "6" K  gSz *,)lFOo- I/IZZdeldmmno05J[^_f[gZhiDN  ($DNEF F)T*IJ J"  z2d;imgsc|jr|dzjn|}|j|}|jj |y)z)Update the state with extracted features.N)rbyterrappend)r5r:rs r8updatezInceptionScore.updates<&*nns   "$>>$' X&r9c t|j}tj|jd}||}|j d}|j d}|j|jd}|j|jd}|Dcgc]}|jdd}}t|||Dcgc]\}}}|||jz z} }}}| D cgc]0} | jdjj2} } tj| } | j| jfScc}wcc}}}wcc} w)zCompute metric.r)dimT)rBkeepdim)r rtorchrandpermshapesoftmax log_softmaxchunkrmeanziplogsumexpstackstd) r5ridxproblog_probp mean_problog_pm_pkl_kkls r8computezInceptionScore.computesB .nnX^^A./C=A&''A'.zz$++1z->>$++1>5;??$QQVV4V0$ ?>> # Example plotting a single value >>> import torch >>> from torchmetrics.image.inception import InceptionScore >>> metric = InceptionScore() >>> metric.update(torch.randint(0, 255, (50, 3, 299, 299), dtype=torch.uint8)) >>> fig_, ax_ = metric.plot() # the returned plot only shows the mean value by default .. plot:: :scale: 75 >>> # Example plotting multiple values >>> import torch >>> from torchmetrics.image.inception import InceptionScore >>> metric = InceptionScore() >>> values = [ ] >>> for _ in range(3): ... # we index by 0 such that only the mean value is plotted ... values.append(metric(torch.randint(0, 255, (50, 3, 299, 299), dtype=torch.uint8))[0]) >>> fig_, ax_ = metric.plot(values) r)r[_plot)r5r\r]s r8plotzInceptionScore.plots+R&T\\^A&zz#r""r9)r" F)NN)__name__ __module__ __qualname____doc__rr3__annotations__rrrfloatlistrrr.rr/rr+rr?tupler[rrrrr` __classcell__)r7s@r8rr#s>@$t#!d!#t#!e!N&OS&,= &<sC'(&<&< &<  &<  &rvsW%'' 5&14[@-.BDTCUVz#Vz#r9