Source code for kornia.losses.lovasz_hinge

import torch
import torch.nn as nn
from torch import Tensor

from kornia.testing import KORNIA_CHECK_SHAPE

# based on:
# https://github.com/bermanmaxim/LovaszSoftmax


[docs]def lovasz_hinge_loss(pred: Tensor, target: Tensor) -> Tensor: r"""Criterion that computes a surrogate binary intersection-over-union (IoU) loss. According to [2], we compute the IoU as follows: .. math:: \text{IoU}(x, class) = \frac{|X \cap Y|}{|X \cup Y|} [1] approximates this fomular with a surrogate, which is fully differentable. Where: - :math:`X` expects to be the scores of each class. - :math:`Y` expects to be the binary tensor with the class labels. the loss, is finally computed as: .. math:: \text{loss}(x, class) = 1 - \text{IoU}(x, class) Reference: [1] http://proceedings.mlr.press/v37/yub15.pdf [2] https://arxiv.org/pdf/1705.08790.pdf . note:: This loss function only supports binary labels. For multi-class labels please use the Lovasz-Softmax loss. Args: pred: logits tensor with shape :math:`(N, 1, H, W)`. labels: labels tensor with shape :math:`(N, H, W)` with binary values. Return: a scalar with the computed loss. Example: >>> N = 1 # num_classes >>> pred = torch.randn(1, N, 3, 5, requires_grad=True) >>> target = torch.empty(1, 3, 5, dtype=torch.long).random_(N) >>> output = lovasz_hinge_loss(pred, target) >>> output.backward() """ KORNIA_CHECK_SHAPE(pred, ["B", "1", "H", "W"]) KORNIA_CHECK_SHAPE(target, ["B", "H", "W"]) if not pred.shape[1] == 1: raise ValueError(f"Invalid pred shape, we expect Bx1xHxW. Got: {pred.shape}") if not pred.shape[-2:] == target.shape[-2:]: raise ValueError(f"pred and target shapes must be the same. Got: {pred.shape} and {target.shape}") if not pred.device == target.device: raise ValueError(f"pred and target must be in the same device. Got: {pred.device} and {target.device}") # flatten pred and target [B, -1] and to float pred_flatten: Tensor = pred.reshape(pred.shape[0], -1) target_flatten: Tensor = target.reshape(target.shape[0], -1) # get shapes B, N = pred_flatten.shape # compute actual loss signs = 2.0 * target_flatten - 1.0 errors = 1.0 - pred_flatten * signs errors_sorted, permutation = errors.sort(dim=1, descending=True) batch_index: Tensor = torch.arange(B, device=pred.device).reshape(-1, 1).repeat(1, N).reshape(-1) target_sorted: Tensor = target_flatten[batch_index, permutation.view(-1)] target_sorted = target_sorted.view(B, N) target_sorted_sum: Tensor = target_sorted.sum(1, keepdim=True) intersection: Tensor = target_sorted_sum - target_sorted.cumsum(1) union: Tensor = target_sorted_sum + (1.0 - target_sorted).cumsum(1) gradient: Tensor = 1.0 - intersection / union if N > 1: gradient[..., 1:] = gradient[..., 1:] - gradient[..., :-1] loss: Tensor = (errors_sorted.relu() * gradient).sum(1).mean() return loss
[docs]class LovaszHingeLoss(nn.Module): r"""Criterion that computes a surrogate binary intersection-over-union (IoU) loss. According to [2], we compute the IoU as follows: .. math:: \text{IoU}(x, class) = \frac{|X \cap Y|}{|X \cup Y|} [1] approximates this fomular with a surrogate, which is fully differentable. Where: - :math:`X` expects to be the scores of each class. - :math:`Y` expects to be the binary tensor with the class labels. the loss, is finally computed as: .. math:: \text{loss}(x, class) = 1 - \text{IoU}(x, class) Reference: [1] http://proceedings.mlr.press/v37/yub15.pdf [2] https://arxiv.org/pdf/1705.08790.pdf . note:: This loss function only supports binary labels. For multi-class labels please use the Lovasz-Softmax loss. Args: pred: logits tensor with shape :math:`(N, 1, H, W)`. labels: labels tensor with shape :math:`(N, H, W)` with binary values. Return: a scalar with the computed loss. Example: >>> N = 1 # num_classes >>> criterion = LovaszHingeLoss() >>> pred = torch.randn(1, N, 3, 5, requires_grad=True) >>> target = torch.empty(1, 3, 5, dtype=torch.long).random_(N) >>> output = criterion(pred, target) >>> output.backward() """ def __init__(self) -> None: super().__init__() def forward(self, pred: Tensor, target: Tensor) -> Tensor: return lovasz_hinge_loss(pred=pred, target=target)