summaryrefslogtreecommitdiff
path: root/src/text_recognizer/datasets/emnist_lines_dataset.py
diff options
context:
space:
mode:
authoraktersnurra <gustaf.rydholm@gmail.com>2020-09-20 11:47:24 +0200
committeraktersnurra <gustaf.rydholm@gmail.com>2020-09-20 11:47:24 +0200
commit1b3b8073a19f939d18a0bb85247eb0d99284f7cc (patch)
treee74e78230ebb179237c063fecf0b52458ce3aa3e /src/text_recognizer/datasets/emnist_lines_dataset.py
parent6137f43c910946301279825e50759a9dd76c6131 (diff)
Bash scripts and some bug fixes.
Diffstat (limited to 'src/text_recognizer/datasets/emnist_lines_dataset.py')
-rw-r--r--src/text_recognizer/datasets/emnist_lines_dataset.py9
1 files changed, 3 insertions, 6 deletions
diff --git a/src/text_recognizer/datasets/emnist_lines_dataset.py b/src/text_recognizer/datasets/emnist_lines_dataset.py
index 6268a01..beb5343 100644
--- a/src/text_recognizer/datasets/emnist_lines_dataset.py
+++ b/src/text_recognizer/datasets/emnist_lines_dataset.py
@@ -149,6 +149,7 @@ class EmnistLinesDataset(Dataset):
# Load emnist dataset.
emnist = EmnistDataset(train=self.train, sample_to_balance=True)
+ emnist.load_or_generate_data()
samples_by_character = get_samples_by_character(
emnist.data.numpy(), emnist.targets.numpy(), self.mapper.mapping,
@@ -306,17 +307,13 @@ def create_datasets(
num_test: int = 1000,
) -> None:
"""Creates a training an validation dataset of Emnist lines."""
- emnist_train = EmnistDataset(train=True, sample_to_balance=True)
- emnist_test = EmnistDataset(train=False, sample_to_balance=True)
- datasets = [emnist_train, emnist_test]
num_samples = [num_train, num_test]
- for num, train, dataset in zip(num_samples, [True, False], datasets):
+ for num, train in zip(num_samples, [True, False]):
emnist_lines = EmnistLinesDataset(
train=train,
- emnist=dataset,
max_length=max_length,
min_overlap=min_overlap,
max_overlap=max_overlap,
num_samples=num,
)
- emnist_lines._load_or_generate_data()
+ emnist_lines.load_or_generate_data()