From 442eac315e4b8be19adab80fb7332d29f68c077c Mon Sep 17 00:00:00 2001 From: Gustaf Rydholm Date: Sun, 27 Jun 2021 20:25:25 +0200 Subject: Fixed bug in word pieces --- text_recognizer/data/iam_paragraphs.py | 4 ++++ 1 file changed, 4 insertions(+) (limited to 'text_recognizer/data/iam_paragraphs.py') diff --git a/text_recognizer/data/iam_paragraphs.py b/text_recognizer/data/iam_paragraphs.py index 6022804..fe60e99 100644 --- a/text_recognizer/data/iam_paragraphs.py +++ b/text_recognizer/data/iam_paragraphs.py @@ -17,6 +17,7 @@ from text_recognizer.data.base_dataset import ( ) from text_recognizer.data.base_data_module import BaseDataModule, load_and_print_info from text_recognizer.data.emnist import emnist_mapping +from text_recognizer.data.mappings import WordPieceMapping from text_recognizer.data.iam import IAM from text_recognizer.data.transforms import WordPiece @@ -49,6 +50,9 @@ class IAMParagraphs(BaseDataModule): self.mapping, self.inverse_mapping, _ = emnist_mapping( extra_symbols=[NEW_LINE_TOKEN] ) + if word_pieces: + self.mapping = WordPieceMapping() + self.train_fraction = train_fraction self.dims = (1, IMAGE_HEIGHT, IMAGE_WIDTH) -- cgit v1.2.3-70-g09d2