kaolin.metrics.voxelgrid

API

kaolin.metrics.voxelgrid.iou(pred, gt)

Computes IoU across two voxelgrids

Parameters:
  • pred (torch.Tensor) – predicted (binary) voxelgrids of shape (N, X, Y, Z)
  • gt (torch.Tensor) – ground-truth (binary) voxelgrids of shape (N, X, Y, Z)
Returns:

the intersection over union value

Return type:

(torch.FloatTensor)

Example

>>> pred = torch.tensor([[[[0., 0.],
...                        [1., 1.]],
...                       [[1., 1.],
...                        [1., 1.]]]])
>>> gt = torch.ones((1,2,2,2))
>>> iou(pred, gt)
tensor([0.7500])