diff options
| -rw-r--r-- | text_recognizer/data/iam_lines.py | 11 |
1 files changed, 8 insertions, 3 deletions
diff --git a/text_recognizer/data/iam_lines.py b/text_recognizer/data/iam_lines.py index 0e45c68..7a063c1 100644 --- a/text_recognizer/data/iam_lines.py +++ b/text_recognizer/data/iam_lines.py @@ -24,8 +24,8 @@ from text_recognizer.data.base_dataset import ( split_dataset, ) from text_recognizer.data.emnist_mapping import EmnistMapping -from text_recognizer.data.iam_paragraphs import get_target_transform from text_recognizer.data.iam import IAM +from text_recognizer.data.iam_paragraphs import get_target_transform ImageFile.LOAD_TRUNCATED_IMAGES = True @@ -35,6 +35,7 @@ PROCESSED_DATA_DIRNAME = BaseDataModule.data_dirname() / "processed" / "iam_line IMAGE_HEIGHT = 56 IMAGE_WIDTH = 1024 MAX_LABEL_LENGTH = 89 +MAX_WORD_PIECE_LENGTH = 72 @attr.s(auto_attribs=True, repr=False) @@ -94,7 +95,9 @@ class IAMLines(BaseDataModule): x_train, y_train, transform=get_transform(IMAGE_WIDTH, self.augment), - target_transform=get_target_transform(self.word_pieces), + target_transform=get_target_transform( + self.word_pieces, max_len=MAX_WORD_PIECE_LENGTH + ), ) self.data_train, self.data_val = split_dataset( @@ -116,7 +119,9 @@ class IAMLines(BaseDataModule): x_test, y_test, transform=get_transform(IMAGE_WIDTH), - target_transform=get_target_transform(self.word_pieces), + target_transform=get_target_transform( + self.word_pieces, max_len=MAX_WORD_PIECE_LENGTH + ), ) if stage is None: |