o .wi%@sddlmZddlmZmZmZddlZddlmZddlm Z ddl m Z m Z m Z ddlmZddlmZdd lmZmZesCd gZGd d d eZdS) )Sequence)AnyOptionalUnionN)Tensor)Literal)_mean_iou_compute_mean_iou_update_mean_iou_validate_args)Metric)_MATPLOTLIB_AVAILABLE)_AX_TYPE_PLOT_OUT_TYPE MeanIoU.plotc seZdZUdZeed<eed<dZeed<dZeed<dZ eed<d Z e ed <d Z e ed <   d"de ededededdedd f fdd Zdededd fddZdefddZd#deeeed fde edefd d!ZZS)$MeanIoUaComputes Mean Intersection over Union (mIoU) for semantic segmentation. The metric is defined by the overlap between the predicted segmentation and the ground truth, divided by the total area covered by the union of the two. The metric can be computed for each class separately or for all classes at once. The metric is optimal at a value of 1 and worst at a value of 0, -1 is returned if class is completely absent both from prediction and the ground truth labels. As input to ``forward`` and ``update`` the metric accepts the following input: - ``preds`` (:class:`~torch.Tensor`): An one-hot boolean tensor of shape ``(N, C, ...)`` with ``N`` being the number of samples and ``C`` the number of classes. Alternatively, an integer tensor of shape ``(N, ...)`` can be provided, where the integer values correspond to the class index. The input type can be controlled with the ``input_format`` argument. - ``target`` (:class:`~torch.Tensor`): An one-hot boolean tensor of shape ``(N, C, ...)`` with ``N`` being the number of samples and ``C`` the number of classes. Alternatively, an integer tensor of shape ``(N, ...)`` can be provided, where the integer values correspond to the class index. The input type can be controlled with the ``input_format`` argument. As output to ``forward`` and ``compute`` the metric returns the following output: - ``miou`` (:class:`~torch.Tensor`): The mean Intersection over Union (mIoU) score. If ``per_class`` is set to ``True``, the output will be a tensor of shape ``(C,)`` with the IoU score for each class. If ``per_class`` is set to ``False``, the output will be a scalar tensor. Args: num_classes: The number of classes in the segmentation problem. Required when input_format="index", optional when input_format="one-hot". include_background: Whether to include the background class in the computation per_class: Whether to compute the IoU for each class separately. If set to ``False``, the metric will compute the mean IoU over all classes. input_format: What kind of input the function receives. Choose between ``"one-hot"`` for one-hot encoded tensors or ``"index"`` for index tensors kwargs: Additional keyword arguments, see :ref:`Metric kwargs` for more info. Raises: ValueError: If ``num_classes`` is not ``None`` or a positive integer ValueError: If ``num_classes`` is not provided when ``input_format="index"`` ValueError: If ``include_background`` is not a boolean ValueError: If ``per_class`` is not a boolean ValueError: If ``input_format`` is not one of ``"one-hot"`` or ``"index"`` Example: >>> import torch >>> from torch import randint >>> from torchmetrics.segmentation import MeanIoU >>> miou = MeanIoU() >>> preds = randint(0, 2, (10, 3, 128, 128), generator=torch.Generator().manual_seed(42)) >>> target = randint(0, 2, (10, 3, 128, 128), generator=torch.Generator().manual_seed(43)) >>> miou(preds, target) tensor(0.3336) >>> miou = MeanIoU(num_classes=3, per_class=True) >>> miou(preds, target) tensor([0.3361, 0.3340, 0.3308]) >>> miou = MeanIoU(per_class=True, include_background=False) >>> miou(preds, target) tensor([0.3340, 0.3308]) >>> miou = MeanIoU(num_classes=3, per_class=True, include_background=True, input_format="index") >>> miou(preds, target) tensor([ 0.3334, 0.3336, -1.0000]) score num_batchesFfull_state_updateis_differentiableThigher_is_betterplot_lower_boundg?plot_upper_boundNone-hot num_classesinclude_background per_class input_format)rindexkwargsreturnc stjdi|t||||||_||_||_||_d|_|durJ|s)|dn|}|jdt |r4|nddd|jdt |ddd|_dS|jdt ddd|jdt ddddS) NFrsumdefaultdist_reduce_fxrT) super__init__r rrrr_is_initialized add_statetorchzeros)selfrrrrr __class__r&_/home/ubuntu/sommelier/.venv/lib/python3.10/site-packages/torchmetrics/segmentation/mean_iou.pyr(js zMeanIoU.__init__predstargetc CsP|js]z|jd|_Wnty }z td|d|d}~ww|jdkr/td|jd|js7|jdn|j}|jdtj||j |j dd d |jd tj||j tj dd d d |_t |||j|j|j \}}t||d d}|dk}|jr|j||jdd7_|j|jdd7_dS|j||7_|j|7_dS)z#Update the state with the new data.r!z4Cannot determine `num_classes` from `preds` tensor: .NrzBExpected argument `num_classes` to be a positive integer, but got r)devicedtyper"r#rTr) zero_division)dim)r)shaper IndexError ValueErrorrr*r+r,r4r5int32r rrrrr"r) r-r1r2errnum_out_classes intersectionunionr valid_classesr&r&r0updatesB  zMeanIoU.updatecCs$|j|j}|jr|dS|S)z6Compute the final Mean Intersection over Union (mIoU).g)rrr nan_to_numnanmean)r- output_scorer&r&r0computes zMeanIoU.computevalaxcCs |||S)abPlot a single or multiple values from the metric. Args: val: Either a single result from calling `metric.forward` or `metric.compute` or a list of these results. If no value is provided, will automatically call `metric.compute` and plot that result. ax: An matplotlib axis object. If provided will add plot to that axis Returns: Figure and Axes object Raises: ModuleNotFoundError: If `matplotlib` is not installed .. plot:: :scale: 75 >>> # Example plotting a single value >>> import torch >>> from torchmetrics.audio import PerceptualEvaluationSpeechQuality >>> metric = PerceptualEvaluationSpeechQuality(8000, 'nb') >>> metric.update(torch.rand(8000), torch.rand(8000)) >>> fig_, ax_ = metric.plot() .. plot:: :scale: 75 >>> # Example plotting multiple values >>> import torch >>> from torchmetrics.audio import PerceptualEvaluationSpeechQuality >>> metric = PerceptualEvaluationSpeechQuality(8000, 'nb') >>> values = [ ] >>> for _ in range(10): ... values.append(metric(torch.rand(8000), torch.rand(8000))) >>> fig_, ax_ = metric.plot(values) )_plot)r-rFrGr&r&r0plots &r)NTFr)NN)__name__ __module__ __qualname____doc__r__annotations__rboolrrrfloatrrintrrr(rArErrr rrI __classcell__r&r&r.r0rs: C     '2r)collections.abcrtypingrrrr+rtyping_extensionsr-torchmetrics.functional.segmentation.mean_iourr r torchmetrics.metricr torchmetrics.utilities.importsr torchmetrics.utilities.plotr r__doctest_skip__rr&r&r&r0s