o .wi` @sPddlmZddlZddlmZmZddlmZddlmZGdddeZ dS))AnyN)Tensortensor)_vif_per_channel)MetriccsreZdZUdZdZdZdZeed<eed<dde de d d ffd d Z d eded d fddZ d efddZ ZS)VisualInformationFidelityuCompute Pixel Based Visual Information Fidelity (VIF_). As input to ``forward`` and ``update`` the metric accepts the following input - ``preds`` (:class:`~torch.Tensor`): Predictions from model of shape ``(N,C,H,W)`` with H,W ≥ 41 - ``target`` (:class:`~torch.Tensor`): Ground truth values of shape ``(N,C,H,W)`` with H,W ≥ 41 As output of `forward` and `compute` the metric returns the following output - ``vif-p`` (:class:`~torch.Tensor`): Tensor with vif-p score Args: sigma_n_sq: variance of the visual noise kwargs: Additional keyword arguments, see :ref:`Metric kwargs` for more info. Example: >>> from torch import randn >>> from torchmetrics.image import VisualInformationFidelity >>> preds = randn([32, 3, 41, 41]) >>> target = randn([32, 3, 41, 41]) >>> vif = VisualInformationFidelity() >>> vif(preds, target) tensor(0.0032) TF vif_scoretotal@ sigma_n_sqkwargsreturnNc s|tjdi|t|tst|tstd||dkr%td||jdtddd|jdtddd||_dS) NzIArgument `sigma_n_sq` is expected to be a positive float or int, but got rrgsum)defaultdist_reduce_fxr ) super__init__ isinstancefloatint ValueError add_staterr )selfr r  __class__rS/home/ubuntu/sommelier/.venv/lib/python3.10/site-packages/torchmetrics/image/vif.pyr9s z"VisualInformationFidelity.__init__predstargetcstd}fddt|D}|dkrtt|dnt|}jt|7_jj d7_dS)z*Update state with predictions and targets.c sJg|]!}tdd|ddddfdd|ddddfjqS)N)rr ).0irrrrr Is>z4VisualInformationFidelity.update..rN) sizerangetorchmeanstackcatrrr shape)rrrchannelsvif_per_channelrr"rupdateFs $z VisualInformationFidelity.updatecCs |j|jS)zCompute vif-p over state.)rr )rrrrcomputePs z!VisualInformationFidelity.compute)r )__name__ __module__ __qualname____doc__is_differentiablehigher_is_betterfull_state_updater__annotations__rrrr-r. __classcell__rrrrrs   r) typingrr&rr!torchmetrics.functional.image.vifrtorchmetrics.metricrrrrrrs