meddlr.ops.categorical_to_one_hot#

meddlr.ops.categorical_to_one_hot(x, channel_dim: int = 1, background=0, num_categories=None, dtype=None)[source]#

Converts categorical predictions to one-hot encoded predictions.

Parameters

  • x (torch.Tensor | np.ndarray) – Categorical array or tensor.

  • channel_dim (int, optional) – Channel dimension for output tensor.

  • background (int | NoneType, optional) – The numerical label of the background category. If None, assumes that the background is a class that should be one-hot encoded.

  • num_categories (int, optional) – Number of categories (excluding background). Defaults to the max(x) + 1.

  • dtype (type, optional) – Data type of the output. Defaults to boolean (torch.bool or np.bool).

Returns

One-hot encoded predictions.

Return type

torch.Tensor | np.ndarray