mindspore.ops.CTCGreedyDecoder

class mindspore.ops.CTCGreedyDecoder(*args, **kwargs)[source]

Performs greedy decoding on the logits given in inputs.

Parameters

merge_repeated (bool) – If true, merge repeated classes in output. Default: True.

Inputs:

  • inputs (Tensor) - The input Tensor must be a 3-D tensor whose shape is (max_time, batch_size, num_classes). num_classes must be num_labels + 1 classes, num_labels indicates the number of actual labels. Blank labels are reserved. Default blank label is num_classes - 1. Data type must be float32 or float64.

  • sequence_length (Tensor) - A tensor containing sequence lengths with the shape of (batch_size). The type must be int32. Each value in the tensor must be equal to or less than max_time.

Outputs:

  • decoded_indices (Tensor) - A tensor with shape of (total_decoded_outputs, 2). Data type is int64.

  • decoded_values (Tensor) - A tensor with shape of (total_decoded_outputs), it stores the decoded classes. Data type is int64.

  • decoded_shape (Tensor) - The value of tensor is [batch_size, max_decoded_legth]. Data type is int64.

  • log_probability (Tensor) - A tensor with shape of (batch_size, 1), containing sequence log-probability, has the same type as inputs.

Raises

  • TypeError – If merge_repeated is not a bool.

  • ValueError – If length of shape of inputs is not equal to 3.

  • ValueError – If length of shape of sequence_length is not equal to 1.

Supported Platforms:

Ascend

Examples

>>> inputs = Tensor(np.random.random((2, 2, 3)), mindspore.float32)
>>> sequence_length = Tensor(np.array([2, 2]), mindspore.int32)
>>> ctc_greedy_decoder = ops.CTCGreedyDecoder()
>>> out1, out2, out3, out4 = ctc_greedy_decoder(inputs, sequence_length)
>>> print(out1, out2, out3, out4)
[[0 0] [0 1] [1 0]]
[0 1 0]
[2 2]
[[-0.7443749] [0.18251707]]