Working on EMNIST Lines dataset

15 files changed, 237 insertions, 0 deletions

diff --git a/tests/__init__.py b/tests/__init__.py
new file mode 100644
index 0000000..18ff212
--- /dev/null
+++ b/tests/__init__.py
@@ -0,0 +1 @@
+"""Test modules for the text text recognizer."""
diff --git a/tests/support/__init__.py b/tests/support/__init__.py
new file mode 100644
index 0000000..a265ede
--- /dev/null
+++ b/tests/support/__init__.py
@@ -0,0 +1,2 @@
+"""Support file modules."""
+from .create_emnist_support_files import create_emnist_support_files
diff --git a/tests/support/create_emnist_lines_support_files.py b/tests/support/create_emnist_lines_support_files.py
new file mode 100644
index 0000000..9abe143
--- /dev/null
+++ b/tests/support/create_emnist_lines_support_files.py
@@ -0,0 +1,51 @@
+"""Module for creating EMNIST Lines test support files."""
+# flake8: noqa: S106
+
+from pathlib import Path
+import shutil
+
+import numpy as np
+
+from text_recognizer.datasets import EmnistLinesDataset
+import text_recognizer.util as util
+
+
+SUPPORT_DIRNAME = Path(__file__).parents[0].resolve() / "emnist_lines"
+
+
+def create_emnist_lines_support_files() -> None:
+    """Create EMNIST Lines test images."""
+    shutil.rmtree(SUPPORT_DIRNAME, ignore_errors=True)
+    SUPPORT_DIRNAME.mkdir()
+
+    # TODO: maybe have to add args to dataset.
+    dataset = EmnistLinesDataset(
+        init_token="<sos>",
+        pad_token="_",
+        eos_token="<eos>",
+        transform=[{"type": "ToTensor", "args": {}}],
+        target_transform=[
+            {
+                "type": "AddTokens",
+                "args": {"init_token": "<sos>", "pad_token": "_", "eos_token": "<eos>"},
+            }
+        ],
+    )  # nosec: S106
+    dataset.load_or_generate_data()
+
+    for index in [5, 7, 9]:
+        image, target = dataset[index]
+        if len(image.shape) == 3:
+            image = image.squeeze(0)
+        print(image.sum(), image.dtype)
+
+        label = "".join(dataset.mapper(label) for label in target[1:]).strip(
+            dataset.mapper.pad_token
+        )
+        print(label)
+        image = image.numpy()
+        util.write_image(image, str(SUPPORT_DIRNAME / f"{label}.png"))
+
+
+if __name__ == "__main__":
+    create_emnist_lines_support_files()
diff --git a/tests/support/create_emnist_support_files.py b/tests/support/create_emnist_support_files.py
new file mode 100644
index 0000000..f9ff030
--- /dev/null
+++ b/tests/support/create_emnist_support_files.py
@@ -0,0 +1,30 @@
+"""Module for creating EMNIST test support files."""
+from pathlib import Path
+import shutil
+
+from text_recognizer.datasets import EmnistDataset
+from text_recognizer.util import write_image
+
+SUPPORT_DIRNAME = Path(__file__).parents[0].resolve() / "emnist"
+
+
+def create_emnist_support_files() -> None:
+    """Create support images for test of CharacterPredictor class."""
+    shutil.rmtree(SUPPORT_DIRNAME, ignore_errors=True)
+    SUPPORT_DIRNAME.mkdir()
+
+    dataset = EmnistDataset(train=False)
+    dataset.load_or_generate_data()
+
+    for index in [5, 7, 9]:
+        image, label = dataset[index]
+        if len(image.shape) == 3:
+            image = image.squeeze(0)
+        image = image.numpy()
+        label = dataset.mapper(int(label))
+        print(index, label)
+        write_image(image, str(SUPPORT_DIRNAME / f"{label}.png"))
+
+
+if __name__ == "__main__":
+    create_emnist_support_files()
diff --git a/tests/support/create_iam_lines_support_files.py b/tests/support/create_iam_lines_support_files.py
new file mode 100644
index 0000000..50f9e3d
--- /dev/null
+++ b/tests/support/create_iam_lines_support_files.py
@@ -0,0 +1,50 @@
+"""Module for creating IAM Lines test support files."""
+# flake8: noqa
+from pathlib import Path
+import shutil
+
+import numpy as np
+
+from text_recognizer.datasets import IamLinesDataset
+import text_recognizer.util as util
+
+
+SUPPORT_DIRNAME = Path(__file__).parents[0].resolve() / "iam_lines"
+
+
+def create_emnist_lines_support_files() -> None:
+    """Create IAM Lines test images."""
+    shutil.rmtree(SUPPORT_DIRNAME, ignore_errors=True)
+    SUPPORT_DIRNAME.mkdir()
+
+    # TODO: maybe have to add args to dataset.
+    dataset = IamLinesDataset(
+        init_token="<sos>",
+        pad_token="_",
+        eos_token="<eos>",
+        transform=[{"type": "ToTensor", "args": {}}],
+        target_transform=[
+            {
+                "type": "AddTokens",
+                "args": {"init_token": "<sos>", "pad_token": "_", "eos_token": "<eos>"},
+            }
+        ],
+    )
+    dataset.load_or_generate_data()
+
+    for index in [0, 1, 3]:
+        image, target = dataset[index]
+        if len(image.shape) == 3:
+            image = image.squeeze(0)
+        print(image.sum(), image.dtype)
+
+        label = "".join(dataset.mapper(label) for label in target[1:]).strip(
+            dataset.mapper.pad_token
+        )
+        print(label)
+        image = image.numpy()
+        util.write_image(image, str(SUPPORT_DIRNAME / f"{label}.png"))
+
+
+if __name__ == "__main__":
+    create_emnist_lines_support_files()
diff --git a/tests/support/emnist_lines/Knox Ky<eos>.png b/tests/support/emnist_lines/Knox Ky<eos>.png
new file mode 100644
index 0000000..b7d0618
--- /dev/null
+++ b/tests/support/emnist_lines/Knox Ky<eos>.png
diff --git a/tests/support/emnist_lines/ancillary beliefs and<eos>.png b/tests/support/emnist_lines/ancillary beliefs and<eos>.png
new file mode 100644
index 0000000..14a8cf3
--- /dev/null
+++ b/tests/support/emnist_lines/ancillary beliefs and<eos>.png
diff --git a/tests/support/emnist_lines/they<eos>.png b/tests/support/emnist_lines/they<eos>.png
new file mode 100644
index 0000000..7f05951
--- /dev/null
+++ b/tests/support/emnist_lines/they<eos>.png
diff --git a/tests/support/iam_lines/He rose from his breakfast-nook bench<eos>.png b/tests/support/iam_lines/He rose from his breakfast-nook bench<eos>.png
new file mode 100644
index 0000000..6eeb642
--- /dev/null
+++ b/tests/support/iam_lines/He rose from his breakfast-nook bench<eos>.png
diff --git a/tests/support/iam_lines/and came into the livingroom, where<eos>.png b/tests/support/iam_lines/and came into the livingroom, where<eos>.png
new file mode 100644
index 0000000..4974cf8
--- /dev/null
+++ b/tests/support/iam_lines/and came into the livingroom, where<eos>.png
diff --git a/tests/support/iam_lines/his entrance. He came, almost falling<eos>.png b/tests/support/iam_lines/his entrance. He came, almost falling<eos>.png
new file mode 100644
index 0000000..a731245
--- /dev/null
+++ b/tests/support/iam_lines/his entrance. He came, almost falling<eos>.png
diff --git a/tests/support/iam_paragraphs/a01-000u.jpg b/tests/support/iam_paragraphs/a01-000u.jpg
new file mode 100644
index 0000000..d9753b6
--- /dev/null
+++ b/tests/support/iam_paragraphs/a01-000u.jpg
diff --git a/tests/test_character_predictor.py b/tests/test_character_predictor.py
new file mode 100644
index 0000000..01bda78
--- /dev/null
+++ b/tests/test_character_predictor.py
@@ -0,0 +1,31 @@
+"""Test for CharacterPredictor class."""
+import importlib
+import os
+from pathlib import Path
+import unittest
+
+from loguru import logger
+
+from text_recognizer.character_predictor import CharacterPredictor
+from text_recognizer.networks import MLP
+
+SUPPORT_DIRNAME = Path(__file__).parents[0].resolve() / "support" / "emnist"
+
+os.environ["CUDA_VISIBLE_DEVICES"] = ""
+
+
+class TestCharacterPredictor(unittest.TestCase):
+    """Tests for the CharacterPredictor class."""
+
+    def test_filename(self) -> None:
+        """Test that CharacterPredictor correctly predicts on a single image, for serveral test images."""
+        network_fn_ = MLP
+        predictor = CharacterPredictor(network_fn=network_fn_)
+
+        for filename in SUPPORT_DIRNAME.glob("*.png"):
+            pred, conf = predictor.predict(str(filename))
+            logger.info(
+                f"Prediction: {pred} at confidence: {conf} for image with character {filename.stem}"
+            )
+            self.assertEqual(pred, filename.stem)
+            self.assertGreater(conf, 0.7)
diff --git a/tests/test_line_predictor.py b/tests/test_line_predictor.py
new file mode 100644
index 0000000..eede4d4
--- /dev/null
+++ b/tests/test_line_predictor.py
@@ -0,0 +1,35 @@
+"""Tests for LinePredictor."""
+import os
+from pathlib import Path
+import unittest
+
+
+import editdistance
+import numpy as np
+
+from text_recognizer.datasets import IamLinesDataset
+from text_recognizer.line_predictor import LinePredictor
+import text_recognizer.util as util
+
+SUPPORT_DIRNAME = Path(__file__).parents[0].resolve() / "support"
+
+os.environ["CUDA_VISIBLE_DEVICES"] = ""
+
+
+class TestEmnistLinePredictor(unittest.TestCase):
+    """Test LinePredictor class on the EmnistLines dataset."""
+
+    def test_filename(self) -> None:
+        """Test that LinePredictor correctly predicts on single images, for several test images."""
+        predictor = LinePredictor(
+            dataset="EmnistLineDataset", network_fn="CNNTransformer"
+        )
+
+        for filename in (SUPPORT_DIRNAME / "emnist_lines").glob("*.png"):
+            pred, conf = predictor.predict(str(filename))
+            true = str(filename.stem)
+            edit_distance = editdistance.eval(pred, true) / len(pred)
+            print(
+                f'Pred: "{pred}" | Confidence: {conf} | True: {true} | Edit distance: {edit_distance}'
+            )
+            self.assertLess(edit_distance, 0.2)
diff --git a/tests/test_paragraph_text_recognizer.py b/tests/test_paragraph_text_recognizer.py
new file mode 100644
index 0000000..3e280b9
--- /dev/null
+++ b/tests/test_paragraph_text_recognizer.py
@@ -0,0 +1,37 @@
+"""Test for ParagraphTextRecognizer class."""
+import os
+from pathlib import Path
+import unittest
+
+from text_recognizer.paragraph_text_recognizer import ParagraphTextRecognizor
+import text_recognizer.util as util
+
+
+SUPPORT_DIRNAME = Path(__file__).parents[0].resolve() / "support" / "iam_paragraph"
+
+# Prevent using GPU.
+os.environ["CUDA_VISIBLE_DEVICES"] = ""
+
+
+class TestParagraphTextRecognizor(unittest.TestCase):
+    """Test that it can take non-square images of max dimension larger than 256px."""
+
+    def test_filename(self) -> None:
+        """Test model on support image."""
+        line_predictor_args = {
+            "dataset": "EmnistLineDataset",
+            "network_fn": "CNNTransformer",
+        }
+        line_detector_args = {"dataset": "EmnistLineDataset", "network_fn": "UNet"}
+        model = ParagraphTextRecognizor(
+            line_predictor_args=line_predictor_args,
+            line_detector_args=line_detector_args,
+        )
+        num_text_lines_by_name = {"a01-000u-cropped": 7}
+        for filename in (SUPPORT_DIRNAME).glob("*.jpg"):
+            full_image = util.read_image(str(filename), grayscale=True)
+            predicted_text, line_region_crops = model.predict(full_image)
+            print(predicted_text)
+            self.assertTrue(
+                len(line_region_crops), num_text_lines_by_name[filename.stem]
+            )