ajmc.ocr.pytorch package

Submodules

ajmc.ocr.pytorch.config module

ajmc.ocr.pytorch.ctc_decoder_torch module

class ajmc.ocr.pytorch.ctc_decoder_torch.BeamCTCDecoder(classes, lm_path=None, alpha=0, beta=0, cutoff_top_n=40, cutoff_prob=1.0, beam_width=100, num_processes=4, blank_index=0)

Bases: Decoder

convert_tensor(offsets, sizes)

convert_to_strings(out, seq_len)

decode(probs, sizes=None)

Decodes probability output using ctcdecode package.

Parameters:

• probs – Tensor of character probabilities, where probs[c,t] is the probability of character c at time t

• sizes – Size of each sequence in the mini-batch

Returns:

sequences of the model’s best guess for the transcription

Return type:

string

class ajmc.ocr.pytorch.ctc_decoder_torch.Decoder(classes, blank_index=0)

Bases: object

Basic decoder class from which all other decoders inherit. Implements several helper functions. Subclasses should implement the decode() method.

Parameters:

• classes (list) – mapping from integers to characters.

• blank_index (int, optional) – index for the blank ‘_’ character. Defaults to 0.

decode(probs, sizes=None)

Given a matrix of character probabilities, returns the decoder’s best guess of the transcription

Parameters:

• probs – Tensor of character probabilities, where probs[c,t] is the probability of character c at time t

• sizes (optional) – Size of each sequence in the mini-batch

Returns:

sequence of the model’s best guess for the transcription

Return type:

string

class ajmc.ocr.pytorch.ctc_decoder_torch.GreedyDecoder(classes, blank_index=0)

Bases: Decoder

decode(probs, sizes=None, remove_repetitions: bool = True) → List[str]

Returns the argmax decoding given the probability matrix. Removes repeated elements in the sequence, as well as blanks.

Parameters:

• probs – Tensor of character probabilities from the network. Expected shape of batch x seq_length x output_dim

• sizes (optional) – Size of each sequence in the mini-batch

• remove_repetitions – Whether to remove repeated characters in the probs

Returns:

sequences of the model’s best guess for the transcription on inputs offsets: time step per character predicted

Return type:

strings

process_string(sequence: torch.tensor, size: int, remove_repetitions=False) → str

ajmc.ocr.pytorch.data_generation module

ajmc.ocr.pytorch.data_processing module

ajmc.ocr.pytorch.datasets module

ajmc.ocr.pytorch.model module

ajmc.ocr.pytorch.train_parallel module