diff --git a/src/ptbench/data/shenzhen/__init__.py b/src/ptbench/data/shenzhen/__init__.py
index 1645962e8cc00443399dd60b88f017c71824e086..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 100644
--- a/src/ptbench/data/shenzhen/__init__.py
+++ b/src/ptbench/data/shenzhen/__init__.py
@@ -1,34 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Shenzhen dataset for computer-aided diagnosis.
-
-The standard digital image database for Tuberculosis is created by the
-National Library of Medicine, Maryland, USA in collaboration with Shenzhen
-No.3 People’s Hospital, Guangdong Medical College, Shenzhen, China.
-The Chest X-rays are from out-patient clinics, and were captured as part of
-the daily routine using Philips DR Digital Diagnose systems.
-
-* Reference: [MONTGOMERY-SHENZHEN-2014]_
-* Original resolution (height x width or width x height): 3000 x 3000 or less
-* Split reference: none
- * Training samples: 64% of TB and healthy CXR (including labels)
- * Validation samples: 16% of TB and healthy CXR (including labels)
- * Test samples: 20% of TB and healthy CXR (including labels)
-"""
-import importlib.resources
-
-_protocols = [
- importlib.resources.files(__name__).joinpath("default.json.bz2"),
- importlib.resources.files(__name__).joinpath("fold_0.json.bz2"),
- importlib.resources.files(__name__).joinpath("fold_1.json.bz2"),
- importlib.resources.files(__name__).joinpath("fold_2.json.bz2"),
- importlib.resources.files(__name__).joinpath("fold_3.json.bz2"),
- importlib.resources.files(__name__).joinpath("fold_4.json.bz2"),
- importlib.resources.files(__name__).joinpath("fold_5.json.bz2"),
- importlib.resources.files(__name__).joinpath("fold_6.json.bz2"),
- importlib.resources.files(__name__).joinpath("fold_7.json.bz2"),
- importlib.resources.files(__name__).joinpath("fold_8.json.bz2"),
- importlib.resources.files(__name__).joinpath("fold_9.json.bz2"),
-]
diff --git a/src/ptbench/data/shenzhen/fold_0.py b/src/ptbench/data/shenzhen/fold_0.py
index 5b4d45602d13a0e6bd0e2724a6c3202c1532eef6..888a0e60024480a3aaff65f6e3d819370fd22669 100644
--- a/src/ptbench/data/shenzhen/fold_0.py
+++ b/src/ptbench/data/shenzhen/fold_0.py
@@ -2,45 +2,46 @@
#
# SPDX-License-Identifier: GPL-3.0-or-later
-"""Shenzhen dataset for TB detection (cross validation fold 0)
+"""Shenzhen datamodule for computer-aided diagnosis (fold 0)
-* Split reference: first 80% of TB and healthy CXR for "train", rest for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.shenzhen` for dataset details
-"""
+See :py:mod:`ptbench.data.shenzhen` for more database details.
-from clapper.logging import setup
+This configuration:
-from .. import return_subsets
-from ..base_datamodule import BaseDataModule
-from . import _maker
+* Raw data input (on disk):
-logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
+ * PNG images (black and white, encoded as color images)
+ * Variable width and height:
+ * widths: from 1130 to 3001 pixels
+ * heights: from 948 to 3001 pixels
-class Fold0Module(BaseDataModule):
- def __init__(
- self,
- train_batch_size=1,
- predict_batch_size=1,
- drop_incomplete_batch=False,
- multiproc_kwargs=None,
- ):
- super().__init__(
- train_batch_size=train_batch_size,
- predict_batch_size=predict_batch_size,
- drop_incomplete_batch=drop_incomplete_batch,
- multiproc_kwargs=multiproc_kwargs,
- )
+* Output image:
+
+ * Transforms:
+
+ * Load raw PNG with :py:mod:`PIL`
+ * Remove black borders
+ * Torch resizing(512px, 512px)
+ * Torch center cropping (512px, 512px)
- def setup(self, stage: str):
- self.dataset = _maker("fold_0")
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
+ * Final specifications:
+ * Fixed resolution: 512x512 pixels
+ * Color RGB encoding
+"""
+
+import importlib.resources
-datamodule = Fold0Module
+from ..datamodule import CachingDataModule
+from ..split import JSONDatabaseSplit
+from .loader import RawDataLoader
+
+datamodule = CachingDataModule(
+ database_split=JSONDatabaseSplit(
+ importlib.resources.files(__name__.rsplit(".", 1)[0]).joinpath(
+ "fold_0.json.bz2"
+ )
+ ),
+ raw_data_loader=RawDataLoader(),
+)
diff --git a/src/ptbench/data/shenzhen/fold_0_rgb.py b/src/ptbench/data/shenzhen/fold_0_rgb.py
deleted file mode 100644
index 143ef731fae0f7d3c746e384e08e400f36c92511..0000000000000000000000000000000000000000
--- a/src/ptbench/data/shenzhen/fold_0_rgb.py
+++ /dev/null
@@ -1,46 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Shenzhen dataset for TB detection (cross validation fold 0, RGB)
-
-* Split reference: first 80% of TB and healthy CXR for "train", rest for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.shenzhen` for dataset details
-"""
-
-from clapper.logging import setup
-
-from .. import return_subsets
-from ..base_datamodule import BaseDataModule
-from . import _maker
-
-logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
-
-
-class Fold0Module(BaseDataModule):
- def __init__(
- self,
- train_batch_size=1,
- predict_batch_size=1,
- drop_incomplete_batch=False,
- multiproc_kwargs=None,
- ):
- super().__init__(
- train_batch_size=train_batch_size,
- predict_batch_size=predict_batch_size,
- drop_incomplete_batch=drop_incomplete_batch,
- multiproc_kwargs=multiproc_kwargs,
- )
-
- def setup(self, stage: str):
- self.dataset = _maker("fold_0", RGB=True)
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
-
-
-datamodule = Fold0Module
diff --git a/src/ptbench/data/shenzhen/fold_1.py b/src/ptbench/data/shenzhen/fold_1.py
index f01adef0af4da152ad756036aacb4bf83c5c20cc..62d7fbd55c83ed746754cbc99dcc65fe48efbc6a 100644
--- a/src/ptbench/data/shenzhen/fold_1.py
+++ b/src/ptbench/data/shenzhen/fold_1.py
@@ -2,45 +2,46 @@
#
# SPDX-License-Identifier: GPL-3.0-or-later
-"""Shenzhen dataset for TB detection (cross validation fold 1)
+"""Shenzhen datamodule for computer-aided diagnosis (fold 1)
-* Split reference: first 80% of TB and healthy CXR for "train", rest for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.shenzhen` for dataset details
-"""
+See :py:mod:`ptbench.data.shenzhen` for more database details.
-from clapper.logging import setup
+This configuration:
-from .. import return_subsets
-from ..base_datamodule import BaseDataModule
-from . import _maker
+* Raw data input (on disk):
-logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
+ * PNG images (black and white, encoded as color images)
+ * Variable width and height:
+ * widths: from 1130 to 3001 pixels
+ * heights: from 948 to 3001 pixels
-class Fold0Module(BaseDataModule):
- def __init__(
- self,
- train_batch_size=1,
- predict_batch_size=1,
- drop_incomplete_batch=False,
- multiproc_kwargs=None,
- ):
- super().__init__(
- train_batch_size=train_batch_size,
- predict_batch_size=predict_batch_size,
- drop_incomplete_batch=drop_incomplete_batch,
- multiproc_kwargs=multiproc_kwargs,
- )
+* Output image:
+
+ * Transforms:
+
+ * Load raw PNG with :py:mod:`PIL`
+ * Remove black borders
+ * Torch resizing(512px, 512px)
+ * Torch center cropping (512px, 512px)
- def setup(self, stage: str):
- self.dataset = _maker("fold_1")
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
+ * Final specifications:
+ * Fixed resolution: 512x512 pixels
+ * Color RGB encoding
+"""
+
+import importlib.resources
-datamodule = Fold0Module
+from ..datamodule import CachingDataModule
+from ..split import JSONDatabaseSplit
+from .loader import RawDataLoader
+
+datamodule = CachingDataModule(
+ database_split=JSONDatabaseSplit(
+ importlib.resources.files(__name__.rsplit(".", 1)[0]).joinpath(
+ "fold_1.json.bz2"
+ )
+ ),
+ raw_data_loader=RawDataLoader(),
+)
diff --git a/src/ptbench/data/shenzhen/fold_1_rgb.py b/src/ptbench/data/shenzhen/fold_1_rgb.py
deleted file mode 100644
index 9d457adfa8835e45c7d5dc993ab23ffc8baafeb9..0000000000000000000000000000000000000000
--- a/src/ptbench/data/shenzhen/fold_1_rgb.py
+++ /dev/null
@@ -1,47 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Shenzhen dataset for TB detection (cross validation fold 1, RGB)
-
-* Split reference: first 80% of TB and healthy CXR for "train", rest for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.shenzhen` for dataset details
-"""
-
-
-from clapper.logging import setup
-
-from .. import return_subsets
-from ..base_datamodule import BaseDataModule
-from . import _maker
-
-logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
-
-
-class Fold0Module(BaseDataModule):
- def __init__(
- self,
- train_batch_size=1,
- predict_batch_size=1,
- drop_incomplete_batch=False,
- multiproc_kwargs=None,
- ):
- super().__init__(
- train_batch_size=train_batch_size,
- predict_batch_size=predict_batch_size,
- drop_incomplete_batch=drop_incomplete_batch,
- multiproc_kwargs=multiproc_kwargs,
- )
-
- def setup(self, stage: str):
- self.dataset = _maker("fold_1", RGB=True)
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
-
-
-datamodule = Fold0Module
diff --git a/src/ptbench/data/shenzhen/fold_2.py b/src/ptbench/data/shenzhen/fold_2.py
index 04dd656263cc6e71d2c61e3f5e221f7129fa7035..b41284cd9d1c4a56c70eff715078f82213dabb3c 100644
--- a/src/ptbench/data/shenzhen/fold_2.py
+++ b/src/ptbench/data/shenzhen/fold_2.py
@@ -2,45 +2,46 @@
#
# SPDX-License-Identifier: GPL-3.0-or-later
-"""Shenzhen dataset for TB detection (cross validation fold 2)
+"""Shenzhen datamodule for computer-aided diagnosis (fold 2)
-* Split reference: first 80% of TB and healthy CXR for "train", rest for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.shenzhen` for dataset details
-"""
+See :py:mod:`ptbench.data.shenzhen` for more database details.
-from clapper.logging import setup
+This configuration:
-from .. import return_subsets
-from ..base_datamodule import BaseDataModule
-from . import _maker
+* Raw data input (on disk):
-logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
+ * PNG images (black and white, encoded as color images)
+ * Variable width and height:
+ * widths: from 1130 to 3001 pixels
+ * heights: from 948 to 3001 pixels
-class Fold0Module(BaseDataModule):
- def __init__(
- self,
- train_batch_size=1,
- predict_batch_size=1,
- drop_incomplete_batch=False,
- multiproc_kwargs=None,
- ):
- super().__init__(
- train_batch_size=train_batch_size,
- predict_batch_size=predict_batch_size,
- drop_incomplete_batch=drop_incomplete_batch,
- multiproc_kwargs=multiproc_kwargs,
- )
+* Output image:
+
+ * Transforms:
+
+ * Load raw PNG with :py:mod:`PIL`
+ * Remove black borders
+ * Torch resizing(512px, 512px)
+ * Torch center cropping (512px, 512px)
- def setup(self, stage: str):
- self.dataset = _maker("fold_2")
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
+ * Final specifications:
+ * Fixed resolution: 512x512 pixels
+ * Color RGB encoding
+"""
+
+import importlib.resources
-datamodule = Fold0Module
+from ..datamodule import CachingDataModule
+from ..split import JSONDatabaseSplit
+from .loader import RawDataLoader
+
+datamodule = CachingDataModule(
+ database_split=JSONDatabaseSplit(
+ importlib.resources.files(__name__.rsplit(".", 1)[0]).joinpath(
+ "fold_2.json.bz2"
+ )
+ ),
+ raw_data_loader=RawDataLoader(),
+)
diff --git a/src/ptbench/data/shenzhen/fold_2_rgb.py b/src/ptbench/data/shenzhen/fold_2_rgb.py
deleted file mode 100644
index 37cbe10ebd72057535bacd241ecb4dbfb2bded3d..0000000000000000000000000000000000000000
--- a/src/ptbench/data/shenzhen/fold_2_rgb.py
+++ /dev/null
@@ -1,46 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Shenzhen dataset for TB detection (cross validation fold 2, RGB)
-
-* Split reference: first 80% of TB and healthy CXR for "train", rest for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.shenzhen` for dataset details
-"""
-
-from clapper.logging import setup
-
-from .. import return_subsets
-from ..base_datamodule import BaseDataModule
-from . import _maker
-
-logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
-
-
-class Fold0Module(BaseDataModule):
- def __init__(
- self,
- train_batch_size=1,
- predict_batch_size=1,
- drop_incomplete_batch=False,
- multiproc_kwargs=None,
- ):
- super().__init__(
- train_batch_size=train_batch_size,
- predict_batch_size=predict_batch_size,
- drop_incomplete_batch=drop_incomplete_batch,
- multiproc_kwargs=multiproc_kwargs,
- )
-
- def setup(self, stage: str):
- self.dataset = _maker("fold_2", RGB=True)
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
-
-
-datamodule = Fold0Module
diff --git a/src/ptbench/data/shenzhen/fold_3.py b/src/ptbench/data/shenzhen/fold_3.py
index b43fcb29c9392d6e9bc5c20d9e93969b33bd2708..cca555064e9923433ef39f591b3e342365cf7afc 100644
--- a/src/ptbench/data/shenzhen/fold_3.py
+++ b/src/ptbench/data/shenzhen/fold_3.py
@@ -2,45 +2,46 @@
#
# SPDX-License-Identifier: GPL-3.0-or-later
-"""Shenzhen dataset for TB detection (cross validation fold 3)
+"""Shenzhen datamodule for computer-aided diagnosis (fold 3)
-* Split reference: first 80% of TB and healthy CXR for "train", rest for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.shenzhen` for dataset details
-"""
+See :py:mod:`ptbench.data.shenzhen` for more database details.
-from clapper.logging import setup
+This configuration:
-from .. import return_subsets
-from ..base_datamodule import BaseDataModule
-from . import _maker
+* Raw data input (on disk):
-logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
+ * PNG images (black and white, encoded as color images)
+ * Variable width and height:
+ * widths: from 1130 to 3001 pixels
+ * heights: from 948 to 3001 pixels
-class Fold0Module(BaseDataModule):
- def __init__(
- self,
- train_batch_size=1,
- predict_batch_size=1,
- drop_incomplete_batch=False,
- multiproc_kwargs=None,
- ):
- super().__init__(
- train_batch_size=train_batch_size,
- predict_batch_size=predict_batch_size,
- drop_incomplete_batch=drop_incomplete_batch,
- multiproc_kwargs=multiproc_kwargs,
- )
+* Output image:
+
+ * Transforms:
+
+ * Load raw PNG with :py:mod:`PIL`
+ * Remove black borders
+ * Torch resizing(512px, 512px)
+ * Torch center cropping (512px, 512px)
- def setup(self, stage: str):
- self.dataset = _maker("fold_3")
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
+ * Final specifications:
+ * Fixed resolution: 512x512 pixels
+ * Color RGB encoding
+"""
+
+import importlib.resources
-datamodule = Fold0Module
+from ..datamodule import CachingDataModule
+from ..split import JSONDatabaseSplit
+from .loader import RawDataLoader
+
+datamodule = CachingDataModule(
+ database_split=JSONDatabaseSplit(
+ importlib.resources.files(__name__.rsplit(".", 1)[0]).joinpath(
+ "fold_3.json.bz2"
+ )
+ ),
+ raw_data_loader=RawDataLoader(),
+)
diff --git a/src/ptbench/data/shenzhen/fold_3_rgb.py b/src/ptbench/data/shenzhen/fold_3_rgb.py
deleted file mode 100644
index 162a3f82d640633ce6d13d25b2a500dc0fb63a54..0000000000000000000000000000000000000000
--- a/src/ptbench/data/shenzhen/fold_3_rgb.py
+++ /dev/null
@@ -1,46 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Shenzhen dataset for TB detection (cross validation fold 3, RGB)
-
-* Split reference: first 80% of TB and healthy CXR for "train", rest for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.shenzhen` for dataset details
-"""
-
-from clapper.logging import setup
-
-from .. import return_subsets
-from ..base_datamodule import BaseDataModule
-from . import _maker
-
-logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
-
-
-class Fold0Module(BaseDataModule):
- def __init__(
- self,
- train_batch_size=1,
- predict_batch_size=1,
- drop_incomplete_batch=False,
- multiproc_kwargs=None,
- ):
- super().__init__(
- train_batch_size=train_batch_size,
- predict_batch_size=predict_batch_size,
- drop_incomplete_batch=drop_incomplete_batch,
- multiproc_kwargs=multiproc_kwargs,
- )
-
- def setup(self, stage: str):
- self.dataset = _maker("fold_3", RGB=True)
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
-
-
-datamodule = Fold0Module
diff --git a/src/ptbench/data/shenzhen/fold_4.py b/src/ptbench/data/shenzhen/fold_4.py
index 58e0a2f221e0a6a48b3466c015c9c5790fc6ae3a..897420076303e47406cc9efb3b6bf0d294ab3611 100644
--- a/src/ptbench/data/shenzhen/fold_4.py
+++ b/src/ptbench/data/shenzhen/fold_4.py
@@ -2,45 +2,46 @@
#
# SPDX-License-Identifier: GPL-3.0-or-later
-"""Shenzhen dataset for TB detection (cross validation fold 4)
+"""Shenzhen datamodule for computer-aided diagnosis (fold 4)
-* Split reference: first 80% of TB and healthy CXR for "train", rest for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.shenzhen` for dataset details
-"""
+See :py:mod:`ptbench.data.shenzhen` for more database details.
-from clapper.logging import setup
+This configuration:
-from .. import return_subsets
-from ..base_datamodule import BaseDataModule
-from . import _maker
+* Raw data input (on disk):
-logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
+ * PNG images (black and white, encoded as color images)
+ * Variable width and height:
+ * widths: from 1130 to 3001 pixels
+ * heights: from 948 to 3001 pixels
-class Fold0Module(BaseDataModule):
- def __init__(
- self,
- train_batch_size=1,
- predict_batch_size=1,
- drop_incomplete_batch=False,
- multiproc_kwargs=None,
- ):
- super().__init__(
- train_batch_size=train_batch_size,
- predict_batch_size=predict_batch_size,
- drop_incomplete_batch=drop_incomplete_batch,
- multiproc_kwargs=multiproc_kwargs,
- )
+* Output image:
+
+ * Transforms:
+
+ * Load raw PNG with :py:mod:`PIL`
+ * Remove black borders
+ * Torch resizing(512px, 512px)
+ * Torch center cropping (512px, 512px)
- def setup(self, stage: str):
- self.dataset = _maker("fold_4")
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
+ * Final specifications:
+ * Fixed resolution: 512x512 pixels
+ * Color RGB encoding
+"""
+
+import importlib.resources
-datamodule = Fold0Module
+from ..datamodule import CachingDataModule
+from ..split import JSONDatabaseSplit
+from .loader import RawDataLoader
+
+datamodule = CachingDataModule(
+ database_split=JSONDatabaseSplit(
+ importlib.resources.files(__name__.rsplit(".", 1)[0]).joinpath(
+ "fold_4.json.bz2"
+ )
+ ),
+ raw_data_loader=RawDataLoader(),
+)
diff --git a/src/ptbench/data/shenzhen/fold_4_rgb.py b/src/ptbench/data/shenzhen/fold_4_rgb.py
deleted file mode 100644
index 0dd4ccf89c553a19030774772447bb66bf0cf9b7..0000000000000000000000000000000000000000
--- a/src/ptbench/data/shenzhen/fold_4_rgb.py
+++ /dev/null
@@ -1,46 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Shenzhen dataset for TB detection (cross validation fold 4, RGB)
-
-* Split reference: first 80% of TB and healthy CXR for "train", rest for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.shenzhen` for dataset details
-"""
-
-from clapper.logging import setup
-
-from .. import return_subsets
-from ..base_datamodule import BaseDataModule
-from . import _maker
-
-logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
-
-
-class Fold0Module(BaseDataModule):
- def __init__(
- self,
- train_batch_size=1,
- predict_batch_size=1,
- drop_incomplete_batch=False,
- multiproc_kwargs=None,
- ):
- super().__init__(
- train_batch_size=train_batch_size,
- predict_batch_size=predict_batch_size,
- drop_incomplete_batch=drop_incomplete_batch,
- multiproc_kwargs=multiproc_kwargs,
- )
-
- def setup(self, stage: str):
- self.dataset = _maker("fold_4", RGB=True)
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
-
-
-datamodule = Fold0Module
diff --git a/src/ptbench/data/shenzhen/fold_5.py b/src/ptbench/data/shenzhen/fold_5.py
index ff115340f723d33ebdeb49524cc7e3b269738563..c520399d98ead9eeb1e3bdcfbe4dc48393adcebc 100644
--- a/src/ptbench/data/shenzhen/fold_5.py
+++ b/src/ptbench/data/shenzhen/fold_5.py
@@ -2,45 +2,46 @@
#
# SPDX-License-Identifier: GPL-3.0-or-later
-"""Shenzhen dataset for TB detection (cross validation fold 5)
+"""Shenzhen datamodule for computer-aided diagnosis (fold 5)
-* Split reference: first 80% of TB and healthy CXR for "train", rest for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.shenzhen` for dataset details
-"""
+See :py:mod:`ptbench.data.shenzhen` for more database details.
-from clapper.logging import setup
+This configuration:
-from .. import return_subsets
-from ..base_datamodule import BaseDataModule
-from . import _maker
+* Raw data input (on disk):
-logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
+ * PNG images (black and white, encoded as color images)
+ * Variable width and height:
+ * widths: from 1130 to 3001 pixels
+ * heights: from 948 to 3001 pixels
-class Fold0Module(BaseDataModule):
- def __init__(
- self,
- train_batch_size=1,
- predict_batch_size=1,
- drop_incomplete_batch=False,
- multiproc_kwargs=None,
- ):
- super().__init__(
- train_batch_size=train_batch_size,
- predict_batch_size=predict_batch_size,
- drop_incomplete_batch=drop_incomplete_batch,
- multiproc_kwargs=multiproc_kwargs,
- )
+* Output image:
+
+ * Transforms:
+
+ * Load raw PNG with :py:mod:`PIL`
+ * Remove black borders
+ * Torch resizing(512px, 512px)
+ * Torch center cropping (512px, 512px)
- def setup(self, stage: str):
- self.dataset = _maker("fold_5")
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
+ * Final specifications:
+ * Fixed resolution: 512x512 pixels
+ * Color RGB encoding
+"""
+
+import importlib.resources
-datamodule = Fold0Module
+from ..datamodule import CachingDataModule
+from ..split import JSONDatabaseSplit
+from .loader import RawDataLoader
+
+datamodule = CachingDataModule(
+ database_split=JSONDatabaseSplit(
+ importlib.resources.files(__name__.rsplit(".", 1)[0]).joinpath(
+ "fold_5.json.bz2"
+ )
+ ),
+ raw_data_loader=RawDataLoader(),
+)
diff --git a/src/ptbench/data/shenzhen/fold_5_rgb.py b/src/ptbench/data/shenzhen/fold_5_rgb.py
deleted file mode 100644
index 46e255e7c37c1f547d25626970324f88d58ba03e..0000000000000000000000000000000000000000
--- a/src/ptbench/data/shenzhen/fold_5_rgb.py
+++ /dev/null
@@ -1,46 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Shenzhen dataset for TB detection (cross validation fold 5, RGB)
-
-* Split reference: first 80% of TB and healthy CXR for "train", rest for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.shenzhen` for dataset details
-"""
-
-from clapper.logging import setup
-
-from .. import return_subsets
-from ..base_datamodule import BaseDataModule
-from . import _maker
-
-logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
-
-
-class Fold0Module(BaseDataModule):
- def __init__(
- self,
- train_batch_size=1,
- predict_batch_size=1,
- drop_incomplete_batch=False,
- multiproc_kwargs=None,
- ):
- super().__init__(
- train_batch_size=train_batch_size,
- predict_batch_size=predict_batch_size,
- drop_incomplete_batch=drop_incomplete_batch,
- multiproc_kwargs=multiproc_kwargs,
- )
-
- def setup(self, stage: str):
- self.dataset = _maker("fold_5", RGB=True)
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
-
-
-datamodule = Fold0Module
diff --git a/src/ptbench/data/shenzhen/fold_6.py b/src/ptbench/data/shenzhen/fold_6.py
index eb81ae882369a1a99bca60e93b1922c7a9d7e9fc..a28f8fc5ca3e0ebd4b49fceaec99d3a2e94dd34c 100644
--- a/src/ptbench/data/shenzhen/fold_6.py
+++ b/src/ptbench/data/shenzhen/fold_6.py
@@ -2,45 +2,46 @@
#
# SPDX-License-Identifier: GPL-3.0-or-later
-"""Shenzhen dataset for TB detection (cross validation fold 6)
+"""Shenzhen datamodule for computer-aided diagnosis (fold 6)
-* Split reference: first 80% of TB and healthy CXR for "train", rest for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.shenzhen` for dataset details
-"""
+See :py:mod:`ptbench.data.shenzhen` for more database details.
-from clapper.logging import setup
+This configuration:
-from .. import return_subsets
-from ..base_datamodule import BaseDataModule
-from . import _maker
+* Raw data input (on disk):
-logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
+ * PNG images (black and white, encoded as color images)
+ * Variable width and height:
+ * widths: from 1130 to 3001 pixels
+ * heights: from 948 to 3001 pixels
-class Fold0Module(BaseDataModule):
- def __init__(
- self,
- train_batch_size=1,
- predict_batch_size=1,
- drop_incomplete_batch=False,
- multiproc_kwargs=None,
- ):
- super().__init__(
- train_batch_size=train_batch_size,
- predict_batch_size=predict_batch_size,
- drop_incomplete_batch=drop_incomplete_batch,
- multiproc_kwargs=multiproc_kwargs,
- )
+* Output image:
+
+ * Transforms:
+
+ * Load raw PNG with :py:mod:`PIL`
+ * Remove black borders
+ * Torch resizing(512px, 512px)
+ * Torch center cropping (512px, 512px)
- def setup(self, stage: str):
- self.dataset = _maker("fold_6")
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
+ * Final specifications:
+ * Fixed resolution: 512x512 pixels
+ * Color RGB encoding
+"""
+
+import importlib.resources
-datamodule = Fold0Module
+from ..datamodule import CachingDataModule
+from ..split import JSONDatabaseSplit
+from .loader import RawDataLoader
+
+datamodule = CachingDataModule(
+ database_split=JSONDatabaseSplit(
+ importlib.resources.files(__name__.rsplit(".", 1)[0]).joinpath(
+ "fold_6.json.bz2"
+ )
+ ),
+ raw_data_loader=RawDataLoader(),
+)
diff --git a/src/ptbench/data/shenzhen/fold_6_rgb.py b/src/ptbench/data/shenzhen/fold_6_rgb.py
deleted file mode 100644
index b9654d08008bd10f0a5d953f2a79870214218651..0000000000000000000000000000000000000000
--- a/src/ptbench/data/shenzhen/fold_6_rgb.py
+++ /dev/null
@@ -1,46 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Shenzhen dataset for TB detection (cross validation fold 6, RGB)
-
-* Split reference: first 80% of TB and healthy CXR for "train", rest for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.shenzhen` for dataset details
-"""
-
-from clapper.logging import setup
-
-from .. import return_subsets
-from ..base_datamodule import BaseDataModule
-from . import _maker
-
-logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
-
-
-class Fold0Module(BaseDataModule):
- def __init__(
- self,
- train_batch_size=1,
- predict_batch_size=1,
- drop_incomplete_batch=False,
- multiproc_kwargs=None,
- ):
- super().__init__(
- train_batch_size=train_batch_size,
- predict_batch_size=predict_batch_size,
- drop_incomplete_batch=drop_incomplete_batch,
- multiproc_kwargs=multiproc_kwargs,
- )
-
- def setup(self, stage: str):
- self.dataset = _maker("fold_6", RGB=True)
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
-
-
-datamodule = Fold0Module
diff --git a/src/ptbench/data/shenzhen/fold_7.py b/src/ptbench/data/shenzhen/fold_7.py
index 79b0d1fff4483ef98eed42b6dce6c6b64076bd1e..b0ea7b4324334980a2e55e4496ac4ab6af705d17 100644
--- a/src/ptbench/data/shenzhen/fold_7.py
+++ b/src/ptbench/data/shenzhen/fold_7.py
@@ -2,45 +2,46 @@
#
# SPDX-License-Identifier: GPL-3.0-or-later
-"""Shenzhen dataset for TB detection (cross validation fold 7)
+"""Shenzhen datamodule for computer-aided diagnosis (fold 7)
-* Split reference: first 80% of TB and healthy CXR for "train", rest for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.shenzhen` for dataset details
-"""
+See :py:mod:`ptbench.data.shenzhen` for more database details.
-from clapper.logging import setup
+This configuration:
-from .. import return_subsets
-from ..base_datamodule import BaseDataModule
-from . import _maker
+* Raw data input (on disk):
-logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
+ * PNG images (black and white, encoded as color images)
+ * Variable width and height:
+ * widths: from 1130 to 3001 pixels
+ * heights: from 948 to 3001 pixels
-class Fold0Module(BaseDataModule):
- def __init__(
- self,
- train_batch_size=1,
- predict_batch_size=1,
- drop_incomplete_batch=False,
- multiproc_kwargs=None,
- ):
- super().__init__(
- train_batch_size=train_batch_size,
- predict_batch_size=predict_batch_size,
- drop_incomplete_batch=drop_incomplete_batch,
- multiproc_kwargs=multiproc_kwargs,
- )
+* Output image:
+
+ * Transforms:
+
+ * Load raw PNG with :py:mod:`PIL`
+ * Remove black borders
+ * Torch resizing(512px, 512px)
+ * Torch center cropping (512px, 512px)
- def setup(self, stage: str):
- self.dataset = _maker("fold_7")
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
+ * Final specifications:
+ * Fixed resolution: 512x512 pixels
+ * Color RGB encoding
+"""
+
+import importlib.resources
-datamodule = Fold0Module
+from ..datamodule import CachingDataModule
+from ..split import JSONDatabaseSplit
+from .loader import RawDataLoader
+
+datamodule = CachingDataModule(
+ database_split=JSONDatabaseSplit(
+ importlib.resources.files(__name__.rsplit(".", 1)[0]).joinpath(
+ "fold_7.json.bz2"
+ )
+ ),
+ raw_data_loader=RawDataLoader(),
+)
diff --git a/src/ptbench/data/shenzhen/fold_7_rgb.py b/src/ptbench/data/shenzhen/fold_7_rgb.py
deleted file mode 100644
index 8a36acb2c79a4467f5dccad532c17cf528613ab5..0000000000000000000000000000000000000000
--- a/src/ptbench/data/shenzhen/fold_7_rgb.py
+++ /dev/null
@@ -1,46 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Shenzhen dataset for TB detection (cross validation fold 7, RGB)
-
-* Split reference: first 80% of TB and healthy CXR for "train", rest for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.shenzhen` for dataset details
-"""
-
-from clapper.logging import setup
-
-from .. import return_subsets
-from ..base_datamodule import BaseDataModule
-from . import _maker
-
-logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
-
-
-class Fold0Module(BaseDataModule):
- def __init__(
- self,
- train_batch_size=1,
- predict_batch_size=1,
- drop_incomplete_batch=False,
- multiproc_kwargs=None,
- ):
- super().__init__(
- train_batch_size=train_batch_size,
- predict_batch_size=predict_batch_size,
- drop_incomplete_batch=drop_incomplete_batch,
- multiproc_kwargs=multiproc_kwargs,
- )
-
- def setup(self, stage: str):
- self.dataset = _maker("fold_7", RGB=True)
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
-
-
-datamodule = Fold0Module
diff --git a/src/ptbench/data/shenzhen/fold_8.py b/src/ptbench/data/shenzhen/fold_8.py
index cf1cd36a8d2d2a40113a775131f9d5b8ba0a092a..9bbfbe84ab942cf5da5a8c5fc8318724908998f9 100644
--- a/src/ptbench/data/shenzhen/fold_8.py
+++ b/src/ptbench/data/shenzhen/fold_8.py
@@ -2,45 +2,46 @@
#
# SPDX-License-Identifier: GPL-3.0-or-later
-"""Shenzhen dataset for TB detection (cross validation fold 8)
+"""Shenzhen datamodule for computer-aided diagnosis (fold 8)
-* Split reference: first 80% of TB and healthy CXR for "train", rest for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.shenzhen` for dataset details
-"""
+See :py:mod:`ptbench.data.shenzhen` for more database details.
-from clapper.logging import setup
+This configuration:
-from .. import return_subsets
-from ..base_datamodule import BaseDataModule
-from . import _maker
+* Raw data input (on disk):
-logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
+ * PNG images (black and white, encoded as color images)
+ * Variable width and height:
+ * widths: from 1130 to 3001 pixels
+ * heights: from 948 to 3001 pixels
-class Fold0Module(BaseDataModule):
- def __init__(
- self,
- train_batch_size=1,
- predict_batch_size=1,
- drop_incomplete_batch=False,
- multiproc_kwargs=None,
- ):
- super().__init__(
- train_batch_size=train_batch_size,
- predict_batch_size=predict_batch_size,
- drop_incomplete_batch=drop_incomplete_batch,
- multiproc_kwargs=multiproc_kwargs,
- )
+* Output image:
+
+ * Transforms:
+
+ * Load raw PNG with :py:mod:`PIL`
+ * Remove black borders
+ * Torch resizing(512px, 512px)
+ * Torch center cropping (512px, 512px)
- def setup(self, stage: str):
- self.dataset = _maker("fold_8")
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
+ * Final specifications:
+ * Fixed resolution: 512x512 pixels
+ * Color RGB encoding
+"""
+
+import importlib.resources
-datamodule = Fold0Module
+from ..datamodule import CachingDataModule
+from ..split import JSONDatabaseSplit
+from .loader import RawDataLoader
+
+datamodule = CachingDataModule(
+ database_split=JSONDatabaseSplit(
+ importlib.resources.files(__name__.rsplit(".", 1)[0]).joinpath(
+ "fold_8.json.bz2"
+ )
+ ),
+ raw_data_loader=RawDataLoader(),
+)
diff --git a/src/ptbench/data/shenzhen/fold_8_rgb.py b/src/ptbench/data/shenzhen/fold_8_rgb.py
deleted file mode 100644
index 1aa0bcec76d875d8e9b966b4461228230bdf79f2..0000000000000000000000000000000000000000
--- a/src/ptbench/data/shenzhen/fold_8_rgb.py
+++ /dev/null
@@ -1,46 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Shenzhen dataset for TB detection (cross validation fold 8, RGB)
-
-* Split reference: first 80% of TB and healthy CXR for "train", rest for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.shenzhen` for dataset details
-"""
-
-from clapper.logging import setup
-
-from .. import return_subsets
-from ..base_datamodule import BaseDataModule
-from . import _maker
-
-logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
-
-
-class Fold0Module(BaseDataModule):
- def __init__(
- self,
- train_batch_size=1,
- predict_batch_size=1,
- drop_incomplete_batch=False,
- multiproc_kwargs=None,
- ):
- super().__init__(
- train_batch_size=train_batch_size,
- predict_batch_size=predict_batch_size,
- drop_incomplete_batch=drop_incomplete_batch,
- multiproc_kwargs=multiproc_kwargs,
- )
-
- def setup(self, stage: str):
- self.dataset = _maker("fold_8", RGB=True)
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
-
-
-datamodule = Fold0Module
diff --git a/src/ptbench/data/shenzhen/fold_9.py b/src/ptbench/data/shenzhen/fold_9.py
index e1bb569d0389a8eb67b0ad186e48c6f98aa7cf57..87c2afb328f9b09f420a1ddce5f5d0ea54346c43 100644
--- a/src/ptbench/data/shenzhen/fold_9.py
+++ b/src/ptbench/data/shenzhen/fold_9.py
@@ -2,45 +2,46 @@
#
# SPDX-License-Identifier: GPL-3.0-or-later
-"""Shenzhen dataset for TB detection (cross validation fold 9)
+"""Shenzhen datamodule for computer-aided diagnosis (fold 9)
-* Split reference: first 80% of TB and healthy CXR for "train", rest for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.shenzhen` for dataset details
-"""
+See :py:mod:`ptbench.data.shenzhen` for more database details.
-from clapper.logging import setup
+This configuration:
-from .. import return_subsets
-from ..base_datamodule import BaseDataModule
-from . import _maker
+* Raw data input (on disk):
-logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
+ * PNG images (black and white, encoded as color images)
+ * Variable width and height:
+ * widths: from 1130 to 3001 pixels
+ * heights: from 948 to 3001 pixels
-class Fold0Module(BaseDataModule):
- def __init__(
- self,
- train_batch_size=1,
- predict_batch_size=1,
- drop_incomplete_batch=False,
- multiproc_kwargs=None,
- ):
- super().__init__(
- train_batch_size=train_batch_size,
- predict_batch_size=predict_batch_size,
- drop_incomplete_batch=drop_incomplete_batch,
- multiproc_kwargs=multiproc_kwargs,
- )
+* Output image:
+
+ * Transforms:
+
+ * Load raw PNG with :py:mod:`PIL`
+ * Remove black borders
+ * Torch resizing(512px, 512px)
+ * Torch center cropping (512px, 512px)
- def setup(self, stage: str):
- self.dataset = _maker("fold_9")
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
+ * Final specifications:
+ * Fixed resolution: 512x512 pixels
+ * Color RGB encoding
+"""
+
+import importlib.resources
-datamodule = Fold0Module
+from ..datamodule import CachingDataModule
+from ..split import JSONDatabaseSplit
+from .loader import RawDataLoader
+
+datamodule = CachingDataModule(
+ database_split=JSONDatabaseSplit(
+ importlib.resources.files(__name__.rsplit(".", 1)[0]).joinpath(
+ "fold_9.json.bz2"
+ )
+ ),
+ raw_data_loader=RawDataLoader(),
+)
diff --git a/src/ptbench/data/shenzhen/fold_9_rgb.py b/src/ptbench/data/shenzhen/fold_9_rgb.py
deleted file mode 100644
index c0a577df0a0d3667420c32042816f65ba9ad20ce..0000000000000000000000000000000000000000
--- a/src/ptbench/data/shenzhen/fold_9_rgb.py
+++ /dev/null
@@ -1,46 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Shenzhen dataset for TB detection (cross validation fold 9, RGB)
-
-* Split reference: first 80% of TB and healthy CXR for "train", rest for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.shenzhen` for dataset details
-"""
-
-from clapper.logging import setup
-
-from .. import return_subsets
-from ..base_datamodule import BaseDataModule
-from . import _maker
-
-logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
-
-
-class Fold0Module(BaseDataModule):
- def __init__(
- self,
- train_batch_size=1,
- predict_batch_size=1,
- drop_incomplete_batch=False,
- multiproc_kwargs=None,
- ):
- super().__init__(
- train_batch_size=train_batch_size,
- predict_batch_size=predict_batch_size,
- drop_incomplete_batch=drop_incomplete_batch,
- multiproc_kwargs=multiproc_kwargs,
- )
-
- def setup(self, stage: str):
- self.dataset = _maker("fold_9", RGB=True)
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
-
-
-datamodule = Fold0Module
diff --git a/src/ptbench/data/shenzhen/rgb.py b/src/ptbench/data/shenzhen/rgb.py
deleted file mode 100644
index 211b49236cae22af68ad61d38849429dedb606d7..0000000000000000000000000000000000000000
--- a/src/ptbench/data/shenzhen/rgb.py
+++ /dev/null
@@ -1,35 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Shenzhen datamodule for computer-aided diagnosis (default protocol)
-
-See :py:mod:`ptbench.data.shenzhen` for dataset details.
-
-This configuration:
-* raw data (default): :py:obj:`ptbench.data.shenzhen._tranforms`
-* augmentations: elastic deformation (probability = 80%)
-* output image resolution: 512x512 pixels
-"""
-
-import importlib.resources
-
-from torchvision import transforms
-
-from ..datamodule import CachingDataModule
-from ..split import JSONDatabaseSplit
-from .loader import RawDataLoader
-
-datamodule = CachingDataModule(
- database_split=JSONDatabaseSplit(
- importlib.resources.files(__name__.rsplit(".", 1)[0]).joinpath(
- "default.json.bz2"
- )
- ),
- raw_data_loader=RawDataLoader(),
- model_transforms=[
- transforms.ToPILImage(),
- transforms.Lambda(lambda x: x.convert("RGB")),
- transforms.ToTensor(),
- ],
-)