Method encode

utils.py:19–38

convert text-label into text-index. input: text: text labels of each image. [batch_size] batch_max_length: max length of text label in the batch. 25 by default output: text: text index for CTCLoss. [batch_size, batch_max_length] length

(self, text, batch_max_length=25)

        self.character = ['[CTCblank]'] + dict_character  # dummy '[CTCblank]' token for CTCLoss (index 0)

    def encode(self, text, batch_max_length=25):
        """convert text-label into text-index.
        input:
            text: text labels of each image. [batch_size]
            batch_max_length: max length of text label in the batch. 25 by default

        output:
            text: text index for CTCLoss. [batch_size, batch_max_length]
            length: length of each text. [batch_size]
        """
        length = [len(s) for s in text]

        # The index used for padding (=0) would not affect the CTC loss calculation.
        batch_text = torch.LongTensor(len(text), batch_max_length).fill_(0)
        for i, t in enumerate(text):
            text = list(t)
            text = [self.dict[char] for char in text]
            # text = [self.dict[char] if char in self.dict.keys() else 0 for char in text]
            batch_text[i][:len(text)] = torch.LongTensor(text)
        return (batch_text.to(device), torch.IntTensor(length).to(device))

    def decode(self, text_index, length):
        """ convert text-index into text-label. """