From bcc1e440d967fffc36d88b9038001415bb1656a9 Mon Sep 17 00:00:00 2001
From: mdelitroz <maxime.delitroz@idiap.ch>
Date: Wed, 2 Aug 2023 16:21:37 +0200
Subject: [PATCH] updated TB-POC dataset and corresponding tests
---
src/ptbench/data/tbpoc/__init__.py | 83 -------------
src/ptbench/data/tbpoc/datamodule.py | 138 +++++++++++++++++++++
src/ptbench/data/tbpoc/fold_0.py | 46 ++-----
src/ptbench/data/tbpoc/fold_1.py | 46 ++-----
src/ptbench/data/tbpoc/fold_2.py | 46 ++-----
src/ptbench/data/tbpoc/fold_3.py | 46 ++-----
src/ptbench/data/tbpoc/fold_4.py | 46 ++-----
src/ptbench/data/tbpoc/fold_5.py | 46 ++-----
src/ptbench/data/tbpoc/fold_6.py | 46 ++-----
src/ptbench/data/tbpoc/fold_7.py | 46 ++-----
src/ptbench/data/tbpoc/fold_8.py | 46 ++-----
src/ptbench/data/tbpoc/fold_9.py | 46 ++-----
tests/test_tbpoc.py | 174 +++++++++++++++------------
13 files changed, 345 insertions(+), 510 deletions(-)
create mode 100644 src/ptbench/data/tbpoc/datamodule.py
diff --git a/src/ptbench/data/tbpoc/__init__.py b/src/ptbench/data/tbpoc/__init__.py
index 00f5f42c..e69de29b 100644
--- a/src/ptbench/data/tbpoc/__init__.py
+++ b/src/ptbench/data/tbpoc/__init__.py
@@ -1,83 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-"""TB-POC dataset for computer-aided diagnosis.
-
-* Reference: [TB-POC-2018]_
-* Original resolution (height x width or width x height): 2048 x 2500
-* Split reference: none
-* Stratified kfold protocol:
-
- * Training samples: 72% of TB and healthy CXR (including labels)
- * Validation samples: 18% of TB and healthy CXR (including labels)
- * Test samples: 10% of TB and healthy CXR (including labels)
-"""
-
-import importlib.resources
-import os
-
-from ...utils.rc import load_rc
-from .. import make_dataset
-from ..dataset import JSONDataset
-from ..loader import load_pil_grayscale, make_delayed
-
-_protocols = [
- 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"),
-]
-
-_datadir = load_rc().get("datadir.tbpoc", os.path.realpath(os.curdir))
-
-
-def _raw_data_loader(sample):
- return dict(
- data=load_pil_grayscale(os.path.join(_datadir, sample["data"])),
- label=sample["label"],
- )
-
-
-def _loader(context, sample):
- # "context" is ignored in this case - database is homogeneous
- # we returned delayed samples to avoid loading all images at once
- return make_delayed(sample, _raw_data_loader)
-
-
-json_dataset = JSONDataset(
- protocols=_protocols,
- fieldnames=("data", "label"),
- loader=_loader,
-)
-"""TB-POC dataset object."""
-
-
-def _maker(protocol, resize_size=512, cc_size=512, RGB=False):
- from torchvision import transforms
-
- from ..augmentations import ElasticDeformation
- from ..image_utils import RemoveBlackBorders
-
- post_transforms = []
- if RGB:
- post_transforms = [
- transforms.Lambda(lambda x: x.convert("RGB")),
- transforms.ToTensor(),
- ]
-
- return make_dataset(
- [json_dataset.subsets(protocol)],
- [
- RemoveBlackBorders(),
- transforms.Resize(resize_size),
- transforms.CenterCrop(cc_size),
- ],
- [ElasticDeformation(p=0.8)],
- post_transforms,
- )
diff --git a/src/ptbench/data/tbpoc/datamodule.py b/src/ptbench/data/tbpoc/datamodule.py
new file mode 100644
index 00000000..35465bac
--- /dev/null
+++ b/src/ptbench/data/tbpoc/datamodule.py
@@ -0,0 +1,138 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+import importlib.resources
+import os
+
+import PIL.Image
+
+from torchvision.transforms.functional import center_crop, to_tensor
+
+from ...utils.rc import load_rc
+from ..datamodule import CachingDataModule
+from ..image_utils import load_pil_grayscale, remove_black_borders
+from ..split import JSONDatabaseSplit
+from ..typing import DatabaseSplit
+from ..typing import RawDataLoader as _BaseRawDataLoader
+from ..typing import Sample
+
+
+class RawDataLoader(_BaseRawDataLoader):
+ """A specialized raw-data-loader for the Shenzen dataset.
+
+ Attributes
+ ----------
+
+ datadir
+ This variable contains the base directory where the database raw data
+ is stored.
+
+ transform
+ Transforms that are always applied to the loaded raw images.
+ """
+
+ datadir: str
+
+ def __init__(self, config_variable: str = "datadir.tbpoc"):
+ self.datadir = load_rc().get(
+ config_variable, os.path.realpath(os.curdir)
+ )
+
+ def sample(self, sample: tuple[str, int]) -> Sample:
+ """Loads a single image sample from the disk.
+
+ Parameters
+ ----------
+
+ sample:
+ A tuple containing the path suffix, within the dataset root folder,
+ where to find the image to be loaded, and an integer, representing the
+ sample label.
+
+
+ Returns
+ -------
+
+ sample
+ The sample representation
+ """
+ image = load_pil_grayscale(os.path.join(self.datadir, sample[0]))
+ image = remove_black_borders(image)
+ tensor = to_tensor(image)
+ tensor = center_crop(tensor, min(*tensor.shape[1:]))
+
+ # use the code below to view generated images
+ # from torchvision.transforms.functional import to_pil_image
+ # to_pil_image(tensor).show()
+ # __import__("pdb").set_trace()
+
+ return tensor, dict(label=sample[1], name=sample[0]) # type: ignore[arg-type]
+
+ def label(self, sample: tuple[str, int]) -> int:
+ """Loads a single image sample label from the disk.
+
+ Parameters
+ ----------
+
+ sample:
+ A tuple containing the path suffix, within the dataset root folder,
+ where to find the image to be loaded, and an integer, representing the
+ sample label.
+
+
+ Returns
+ -------
+
+ label
+ The integer label associated with the sample
+ """
+ return sample[1]
+
+
+def make_split(basename: str) -> DatabaseSplit:
+ """Returns a database split for the TB-POC database."""
+
+ return JSONDatabaseSplit(
+ importlib.resources.files(__name__.rsplit(".", 1)[0]).joinpath(basename)
+ )
+
+
+class DataModule(CachingDataModule):
+ """TB-POC dataset for computer-aided diagnosis.
+
+ * Database reference: [TB-POC-2018]_
+ * Original resolution (height x width or width x height): 2048 x 2500 pixels
+ or 2500 x 2048 pixels
+
+ Data specifications:
+
+ * Raw data input (on disk):
+
+ * jpeg 8-bit grayscale images
+ * resolution: fixed to one of the cases above
+
+ * Output image:
+
+ * Transforms:
+
+ * Load raw jpeg with :py:mod:`PIL`
+ * Remove black borders
+ * Convert to torch tensor
+ * Torch center cropping to get square image
+
+ * Final specifications:
+
+ * Grayscale, encoded as a single plane tensor, 32-bit floats,
+ square with varying resolutions, depending on black borders' sizes
+ on the input image
+ * Labels: 0 (healthy), 1 (active tuberculosis)
+ """
+
+ def __init__(self, split_filename: str):
+ super().__init__(
+ database_split=make_split(split_filename),
+ raw_data_loader=RawDataLoader(),
+ )
+
+
diff --git a/src/ptbench/data/tbpoc/fold_0.py b/src/ptbench/data/tbpoc/fold_0.py
index 7a423deb..972e7188 100644
--- a/src/ptbench/data/tbpoc/fold_0.py
+++ b/src/ptbench/data/tbpoc/fold_0.py
@@ -1,45 +1,21 @@
# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
#
# SPDX-License-Identifier: GPL-3.0-or-later
+
"""TB-POC dataset for TB detection (cross validation fold 0)
* Split reference: none (stratified kfolding)
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbpoc` 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")
+* Stratified kfold protocol:
+ * Training samples: 72% of TB and healthy CXR (including labels)
+ * Validation samples: 18% of TB and healthy CXR (including labels)
+ * Test samples: 10% of TB and healthy CXR (including labels)
+* This configuration resolution: varying depending of black borders on original
+ image
+* See :py:mod:`ptbench.data.tbpoc` for dataset details
+"""
-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")
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
-
+from .datamodule import DataModule
-datamodule = Fold0Module
+datamodule = DataModule("fold-0.json")
diff --git a/src/ptbench/data/tbpoc/fold_1.py b/src/ptbench/data/tbpoc/fold_1.py
index cb4c59ba..79b9bfca 100644
--- a/src/ptbench/data/tbpoc/fold_1.py
+++ b/src/ptbench/data/tbpoc/fold_1.py
@@ -1,45 +1,21 @@
# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
#
# SPDX-License-Identifier: GPL-3.0-or-later
+
"""TB-POC dataset for TB detection (cross validation fold 1)
* Split reference: none (stratified kfolding)
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbpoc` 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")
+* Stratified kfold protocol:
+ * Training samples: 72% of TB and healthy CXR (including labels)
+ * Validation samples: 18% of TB and healthy CXR (including labels)
+ * Test samples: 10% of TB and healthy CXR (including labels)
+* This configuration resolution: varying depending of black borders on original
+ image
+* See :py:mod:`ptbench.data.tbpoc` for dataset details
+"""
-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")
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
-
+from .datamodule import DataModule
-datamodule = Fold0Module
+datamodule = DataModule("fold-1.json")
diff --git a/src/ptbench/data/tbpoc/fold_2.py b/src/ptbench/data/tbpoc/fold_2.py
index 1bffecea..9d41fb59 100644
--- a/src/ptbench/data/tbpoc/fold_2.py
+++ b/src/ptbench/data/tbpoc/fold_2.py
@@ -1,45 +1,21 @@
# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
#
# SPDX-License-Identifier: GPL-3.0-or-later
+
"""TB-POC dataset for TB detection (cross validation fold 2)
* Split reference: none (stratified kfolding)
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbpoc` 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")
+* Stratified kfold protocol:
+ * Training samples: 72% of TB and healthy CXR (including labels)
+ * Validation samples: 18% of TB and healthy CXR (including labels)
+ * Test samples: 10% of TB and healthy CXR (including labels)
+* This configuration resolution: varying depending of black borders on original
+ image
+* See :py:mod:`ptbench.data.tbpoc` for dataset details
+"""
-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")
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
-
+from .datamodule import DataModule
-datamodule = Fold0Module
+datamodule = DataModule("fold-2.json")
diff --git a/src/ptbench/data/tbpoc/fold_3.py b/src/ptbench/data/tbpoc/fold_3.py
index 1263d39b..08672b3f 100644
--- a/src/ptbench/data/tbpoc/fold_3.py
+++ b/src/ptbench/data/tbpoc/fold_3.py
@@ -1,45 +1,21 @@
# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
#
# SPDX-License-Identifier: GPL-3.0-or-later
+
"""TB-POC dataset for TB detection (cross validation fold 3)
* Split reference: none (stratified kfolding)
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbpoc` 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")
+* Stratified kfold protocol:
+ * Training samples: 72% of TB and healthy CXR (including labels)
+ * Validation samples: 18% of TB and healthy CXR (including labels)
+ * Test samples: 10% of TB and healthy CXR (including labels)
+* This configuration resolution: varying depending of black borders on original
+ image
+* See :py:mod:`ptbench.data.tbpoc` for dataset details
+"""
-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")
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
-
+from .datamodule import DataModule
-datamodule = Fold0Module
+datamodule = DataModule("fold-3.json")
diff --git a/src/ptbench/data/tbpoc/fold_4.py b/src/ptbench/data/tbpoc/fold_4.py
index 119adfa9..8354a4c2 100644
--- a/src/ptbench/data/tbpoc/fold_4.py
+++ b/src/ptbench/data/tbpoc/fold_4.py
@@ -1,45 +1,21 @@
# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
#
# SPDX-License-Identifier: GPL-3.0-or-later
+
"""TB-POC dataset for TB detection (cross validation fold 4)
* Split reference: none (stratified kfolding)
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbpoc` 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")
+* Stratified kfold protocol:
+ * Training samples: 72% of TB and healthy CXR (including labels)
+ * Validation samples: 18% of TB and healthy CXR (including labels)
+ * Test samples: 10% of TB and healthy CXR (including labels)
+* This configuration resolution: varying depending of black borders on original
+ image
+* See :py:mod:`ptbench.data.tbpoc` for dataset details
+"""
-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")
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
-
+from .datamodule import DataModule
-datamodule = Fold0Module
+datamodule = DataModule("fold-4.json")
diff --git a/src/ptbench/data/tbpoc/fold_5.py b/src/ptbench/data/tbpoc/fold_5.py
index 2a90cbdb..cb7f9561 100644
--- a/src/ptbench/data/tbpoc/fold_5.py
+++ b/src/ptbench/data/tbpoc/fold_5.py
@@ -1,45 +1,21 @@
# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
#
# SPDX-License-Identifier: GPL-3.0-or-later
+
"""TB-POC dataset for TB detection (cross validation fold 5)
* Split reference: none (stratified kfolding)
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbpoc` 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")
+* Stratified kfold protocol:
+ * Training samples: 72% of TB and healthy CXR (including labels)
+ * Validation samples: 18% of TB and healthy CXR (including labels)
+ * Test samples: 10% of TB and healthy CXR (including labels)
+* This configuration resolution: varying depending of black borders on original
+ image
+* See :py:mod:`ptbench.data.tbpoc` for dataset details
+"""
-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")
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
-
+from .datamodule import DataModule
-datamodule = Fold0Module
+datamodule = DataModule("fold-5.json")
diff --git a/src/ptbench/data/tbpoc/fold_6.py b/src/ptbench/data/tbpoc/fold_6.py
index 42ed763d..379211aa 100644
--- a/src/ptbench/data/tbpoc/fold_6.py
+++ b/src/ptbench/data/tbpoc/fold_6.py
@@ -1,45 +1,21 @@
# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
#
# SPDX-License-Identifier: GPL-3.0-or-later
+
"""TB-POC dataset for TB detection (cross validation fold 6)
* Split reference: none (stratified kfolding)
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbpoc` 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")
+* Stratified kfold protocol:
+ * Training samples: 72% of TB and healthy CXR (including labels)
+ * Validation samples: 18% of TB and healthy CXR (including labels)
+ * Test samples: 10% of TB and healthy CXR (including labels)
+* This configuration resolution: varying depending of black borders on original
+ image
+* See :py:mod:`ptbench.data.tbpoc` for dataset details
+"""
-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")
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
-
+from .datamodule import DataModule
-datamodule = Fold0Module
+datamodule = DataModule("fold-6.json")
diff --git a/src/ptbench/data/tbpoc/fold_7.py b/src/ptbench/data/tbpoc/fold_7.py
index ad7dbe14..b846b88a 100644
--- a/src/ptbench/data/tbpoc/fold_7.py
+++ b/src/ptbench/data/tbpoc/fold_7.py
@@ -1,45 +1,21 @@
# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
#
# SPDX-License-Identifier: GPL-3.0-or-later
+
"""TB-POC dataset for TB detection (cross validation fold 7)
* Split reference: none (stratified kfolding)
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbpoc` 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")
+* Stratified kfold protocol:
+ * Training samples: 72% of TB and healthy CXR (including labels)
+ * Validation samples: 18% of TB and healthy CXR (including labels)
+ * Test samples: 10% of TB and healthy CXR (including labels)
+* This configuration resolution: varying depending of black borders on original
+ image
+* See :py:mod:`ptbench.data.tbpoc` for dataset details
+"""
-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")
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
-
+from .datamodule import DataModule
-datamodule = Fold0Module
+datamodule = DataModule("fold-7.json")
diff --git a/src/ptbench/data/tbpoc/fold_8.py b/src/ptbench/data/tbpoc/fold_8.py
index 4bcea788..acfd4296 100644
--- a/src/ptbench/data/tbpoc/fold_8.py
+++ b/src/ptbench/data/tbpoc/fold_8.py
@@ -1,45 +1,21 @@
# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
#
# SPDX-License-Identifier: GPL-3.0-or-later
+
"""TB-POC dataset for TB detection (cross validation fold 8)
* Split reference: none (stratified kfolding)
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbpoc` 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")
+* Stratified kfold protocol:
+ * Training samples: 72% of TB and healthy CXR (including labels)
+ * Validation samples: 18% of TB and healthy CXR (including labels)
+ * Test samples: 10% of TB and healthy CXR (including labels)
+* This configuration resolution: varying depending of black borders on original
+ image
+* See :py:mod:`ptbench.data.tbpoc` for dataset details
+"""
-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")
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
-
+from .datamodule import DataModule
-datamodule = Fold0Module
+datamodule = DataModule("fold-8.json")
diff --git a/src/ptbench/data/tbpoc/fold_9.py b/src/ptbench/data/tbpoc/fold_9.py
index c33eb6bd..4634068e 100644
--- a/src/ptbench/data/tbpoc/fold_9.py
+++ b/src/ptbench/data/tbpoc/fold_9.py
@@ -1,45 +1,21 @@
# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
#
# SPDX-License-Identifier: GPL-3.0-or-later
+
"""TB-POC dataset for TB detection (cross validation fold 9)
* Split reference: none (stratified kfolding)
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbpoc` 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")
+* Stratified kfold protocol:
+ * Training samples: 72% of TB and healthy CXR (including labels)
+ * Validation samples: 18% of TB and healthy CXR (including labels)
+ * Test samples: 10% of TB and healthy CXR (including labels)
+* This configuration resolution: varying depending of black borders on original
+ image
+* See :py:mod:`ptbench.data.tbpoc` for dataset details
+"""
-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")
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
-
+from .datamodule import DataModule
-datamodule = Fold0Module
+datamodule = DataModule("fold-9.json")
diff --git a/tests/test_tbpoc.py b/tests/test_tbpoc.py
index 9609ea66..ee34d8d0 100644
--- a/tests/test_tbpoc.py
+++ b/tests/test_tbpoc.py
@@ -4,106 +4,126 @@
"""Tests for TB-POC dataset."""
import pytest
+import torch
-dataset = None
+from ptbench.data.tbpoc.datamodule import make_split
-@pytest.mark.skip(reason="Test need to be updated")
-def test_protocol_consistency():
- # Cross-validation fold 0-6
- for f in range(7):
- subset = dataset.subsets("fold_" + str(f))
- assert len(subset) == 3
+def _check_split(
+ split_filename: str,
+ lengths: dict[str, int],
+ prefix: str = "TBPOC_CXR/",
+ extension: str = ".jpeg",
+ possible_labels: list[int] = [0, 1],
+):
+ """Runs a simple consistence check on the data split.
- assert "train" in subset
- assert len(subset["train"]) == 292
- for s in subset["train"]:
- assert s.key.upper().startswith("TBPOC_CXR/TBPOC-")
+ Parameters
+ ----------
- assert "validation" in subset
- assert len(subset["validation"]) == 74
- for s in subset["validation"]:
- assert s.key.upper().startswith("TBPOC_CXR/TBPOC-")
+ split_filename
+ This is the split we will check
- assert "test" in subset
- assert len(subset["test"]) == 41
- for s in subset["test"]:
- assert s.key.upper().startswith("TBPOC_CXR/TBPOC-")
+ lenghts
+ A dictionary that contains keys matching those of the split (this will
+ be checked). The values of the dictionary should correspond to the
+ sizes of each of the datasets in the split.
- # Check labels
- for s in subset["train"]:
- assert s.label in [0.0, 1.0]
+ prefix
+ Each file named in a split should start with this prefix.
- for s in subset["validation"]:
- assert s.label in [0.0, 1.0]
+ extension
+ Each file named in a split should end with this extension.
- for s in subset["test"]:
- assert s.label in [0.0, 1.0]
+ possible_labels
+ These are the list of possible labels contained in any split.
+ """
- # Cross-validation fold 7-9
- for f in range(7, 10):
- subset = dataset.subsets("fold_" + str(f))
- assert len(subset) == 3
+ split = make_split(split_filename)
- assert "train" in subset
- assert len(subset["train"]) == 293
- for s in subset["train"]:
- assert s.key.upper().startswith("TBPOC_CXR/TBPOC-")
+ assert len(split) == len(lengths)
- assert "validation" in subset
- assert len(subset["validation"]) == 74
- for s in subset["validation"]:
- assert s.key.upper().startswith("TBPOC_CXR/TBPOC-")
+ for k in lengths.keys():
+ # dataset must have been declared
+ assert k in split
- assert "test" in subset
- assert len(subset["test"]) == 40
- for s in subset["test"]:
- assert s.key.upper().startswith("TBPOC_CXR/TBPOC-")
+ assert len(split[k]) == lengths[k]
+ for s in split[k]:
+ # assert s[0].startswith(prefix)
+ assert s[0].endswith(extension)
+ assert s[1] in possible_labels
- # Check labels
- for s in subset["train"]:
- assert s.label in [0.0, 1.0]
- for s in subset["validation"]:
- assert s.label in [0.0, 1.0]
+def _check_loaded_batch(
+ batch,
+ size: int = 1,
+ prefix: str = "TBPOC_CXR/",
+ extension: str = ".jpeg",
+ possible_labels: list[int] = [0, 1],
+):
+ """Checks the consistence of an individual (loaded) batch.
- for s in subset["test"]:
- assert s.label in [0.0, 1.0]
+ Parameters
+ ----------
+ batch
+ The loaded batch to be checked.
-@pytest.mark.skip(reason="Test need to be updated")
-@pytest.mark.skip_if_rc_var_not_set("datadir.tbpoc")
-def test_loading():
- image_size_portrait = (2048, 2500)
- image_size_landscape = (2500, 2048)
+ prefix
+ Each file named in a split should start with this prefix.
+
+ extension
+ Each file named in a split should end with this extension.
+
+ possible_labels
+ These are the list of possible labels contained in any split.
+ """
- def _check_size(size):
- if size == image_size_portrait:
- return True
- elif size == image_size_landscape:
- return True
- return False
+ assert len(batch) == 2 # data, metadata
- def _check_sample(s):
- data = s.data
- assert isinstance(data, dict)
- assert len(data) == 2
+ assert isinstance(batch[0], torch.Tensor)
+ assert batch[0].shape[0] == size # mini-batch size
+ assert batch[0].shape[1] == 1 # grayscale images
+ assert batch[0].shape[2] == batch[0].shape[3] # image is square
- assert "data" in data
- assert _check_size(data["data"].size) # Check size
- assert data["data"].mode, "L" # Check colors
+ assert isinstance(batch[1], dict) # metadata
+ assert len(batch[1]) == 2 # label and name
- assert "label" in data
- assert data["label"] in [0, 1] # Check labels
+ assert "label" in batch[1]
+ assert all([k in possible_labels for k in batch[1]["label"]])
- limit = 30 # use this to limit testing to first images only, else None
+ assert "name" in batch[1]
+ # assert all([k.startswith(prefix) for k in batch[1]["name"]])
+ assert all([k.endswith(extension) for k in batch[1]["name"]])
- subset = dataset.subsets("fold_0")
- for s in subset["train"][:limit]:
- _check_sample(s)
+
+def test_protocol_consistency():
+ # Cross-validation fold 0-6
+ for k in range(7):
+ _check_split(
+ f"fold-{k}.json",
+ lengths=dict(train=292, validation=74, test=41),
+ )
+
+ # Cross-validation fold 7-9
+ for k in range(7, 10):
+ _check_split(
+ f"fold-{k}.json",
+ lengths=dict(train=293, validation=74, test=40),
+ )
-@pytest.mark.skip(reason="Test need to be updated")
-@pytest.mark.skip_if_rc_var_not_set("datadir.tbpoc")
-def test_check():
- assert dataset.check() == 0
+@pytest.mark.skip_if_rc_var_not_set("datadir.hivtb")
+def test_loading():
+ from ptbench.data.tbpoc.fold_0 import datamodule
+
+ datamodule.model_transforms = [] # should be done before setup()
+ datamodule.setup("predict") # sets up all datasets
+
+ for loader in datamodule.predict_dataloader().values():
+ limit = 5 # limit load checking
+ for batch in loader:
+ if limit == 0:
+ break
+ _check_loaded_batch(batch)
+ limit -= 1
--
GitLab