“Index Pytorch Jaccard” Réponses codées

Index Pytorch Jaccard

>>> from torchmetrics import JaccardIndex
>>> target = torch.randint(0, 2, (10, 25, 25))
>>> pred = torch.tensor(target)
>>> pred[2:5, 7:13, 9:15] = 1 - pred[2:5, 7:13, 9:15]
>>> jaccard = JaccardIndex(num_classes=2)
>>> jaccard(pred, target)
tensor(0.9660)