From fb3e851945a527cfec2c358f06e79ebf0f5c18f0 Mon Sep 17 00:00:00 2001
From: dcarron <daniel.carron@idiap.ch>
Date: Thu, 1 Jun 2023 15:22:41 +0200
Subject: [PATCH] Moved montgomery configs to data
---
pyproject.toml | 44 ++++++++---------
.../configs/datasets/montgomery/__init__.py | 29 ------------
.../configs/datasets/montgomery/default.py | 15 ------
.../configs/datasets/montgomery/fold_0.py | 15 ------
.../configs/datasets/montgomery/fold_0_rgb.py | 15 ------
.../configs/datasets/montgomery/fold_1.py | 15 ------
.../configs/datasets/montgomery/fold_1_rgb.py | 15 ------
.../configs/datasets/montgomery/fold_2.py | 15 ------
.../configs/datasets/montgomery/fold_2_rgb.py | 15 ------
.../configs/datasets/montgomery/fold_3.py | 15 ------
.../configs/datasets/montgomery/fold_3_rgb.py | 15 ------
.../configs/datasets/montgomery/fold_4.py | 15 ------
.../configs/datasets/montgomery/fold_4_rgb.py | 15 ------
.../configs/datasets/montgomery/fold_5.py | 15 ------
.../configs/datasets/montgomery/fold_5_rgb.py | 15 ------
.../configs/datasets/montgomery/fold_6.py | 15 ------
.../configs/datasets/montgomery/fold_6_rgb.py | 15 ------
.../configs/datasets/montgomery/fold_7.py | 15 ------
.../configs/datasets/montgomery/fold_7_rgb.py | 16 -------
.../configs/datasets/montgomery/fold_8.py | 15 ------
.../configs/datasets/montgomery/fold_8_rgb.py | 15 ------
.../configs/datasets/montgomery/fold_9.py | 15 ------
.../configs/datasets/montgomery/fold_9_rgb.py | 15 ------
.../configs/datasets/montgomery/rgb.py | 15 ------
src/ptbench/data/montgomery/__init__.py | 27 ++++++++++-
src/ptbench/data/montgomery/default.py | 47 +++++++++++++++++++
src/ptbench/data/montgomery/fold_0.py | 47 +++++++++++++++++++
src/ptbench/data/montgomery/fold_0_rgb.py | 47 +++++++++++++++++++
src/ptbench/data/montgomery/fold_1.py | 47 +++++++++++++++++++
src/ptbench/data/montgomery/fold_1_rgb.py | 47 +++++++++++++++++++
src/ptbench/data/montgomery/fold_2.py | 47 +++++++++++++++++++
src/ptbench/data/montgomery/fold_2_rgb.py | 47 +++++++++++++++++++
src/ptbench/data/montgomery/fold_3.py | 47 +++++++++++++++++++
src/ptbench/data/montgomery/fold_3_rgb.py | 47 +++++++++++++++++++
src/ptbench/data/montgomery/fold_4.py | 47 +++++++++++++++++++
src/ptbench/data/montgomery/fold_4_rgb.py | 47 +++++++++++++++++++
src/ptbench/data/montgomery/fold_5.py | 47 +++++++++++++++++++
src/ptbench/data/montgomery/fold_5_rgb.py | 47 +++++++++++++++++++
src/ptbench/data/montgomery/fold_6.py | 47 +++++++++++++++++++
src/ptbench/data/montgomery/fold_6_rgb.py | 47 +++++++++++++++++++
src/ptbench/data/montgomery/fold_7.py | 47 +++++++++++++++++++
src/ptbench/data/montgomery/fold_7_rgb.py | 47 +++++++++++++++++++
src/ptbench/data/montgomery/fold_8.py | 47 +++++++++++++++++++
src/ptbench/data/montgomery/fold_8_rgb.py | 47 +++++++++++++++++++
src/ptbench/data/montgomery/fold_9.py | 47 +++++++++++++++++++
src/ptbench/data/montgomery/fold_9_rgb.py | 47 +++++++++++++++++++
src/ptbench/data/montgomery/rgb.py | 47 +++++++++++++++++++
47 files changed, 1082 insertions(+), 383 deletions(-)
delete mode 100644 src/ptbench/configs/datasets/montgomery/__init__.py
delete mode 100644 src/ptbench/configs/datasets/montgomery/default.py
delete mode 100644 src/ptbench/configs/datasets/montgomery/fold_0.py
delete mode 100644 src/ptbench/configs/datasets/montgomery/fold_0_rgb.py
delete mode 100644 src/ptbench/configs/datasets/montgomery/fold_1.py
delete mode 100644 src/ptbench/configs/datasets/montgomery/fold_1_rgb.py
delete mode 100644 src/ptbench/configs/datasets/montgomery/fold_2.py
delete mode 100644 src/ptbench/configs/datasets/montgomery/fold_2_rgb.py
delete mode 100644 src/ptbench/configs/datasets/montgomery/fold_3.py
delete mode 100644 src/ptbench/configs/datasets/montgomery/fold_3_rgb.py
delete mode 100644 src/ptbench/configs/datasets/montgomery/fold_4.py
delete mode 100644 src/ptbench/configs/datasets/montgomery/fold_4_rgb.py
delete mode 100644 src/ptbench/configs/datasets/montgomery/fold_5.py
delete mode 100644 src/ptbench/configs/datasets/montgomery/fold_5_rgb.py
delete mode 100644 src/ptbench/configs/datasets/montgomery/fold_6.py
delete mode 100644 src/ptbench/configs/datasets/montgomery/fold_6_rgb.py
delete mode 100644 src/ptbench/configs/datasets/montgomery/fold_7.py
delete mode 100644 src/ptbench/configs/datasets/montgomery/fold_7_rgb.py
delete mode 100644 src/ptbench/configs/datasets/montgomery/fold_8.py
delete mode 100644 src/ptbench/configs/datasets/montgomery/fold_8_rgb.py
delete mode 100644 src/ptbench/configs/datasets/montgomery/fold_9.py
delete mode 100644 src/ptbench/configs/datasets/montgomery/fold_9_rgb.py
delete mode 100644 src/ptbench/configs/datasets/montgomery/rgb.py
create mode 100644 src/ptbench/data/montgomery/default.py
create mode 100644 src/ptbench/data/montgomery/fold_0.py
create mode 100644 src/ptbench/data/montgomery/fold_0_rgb.py
create mode 100644 src/ptbench/data/montgomery/fold_1.py
create mode 100644 src/ptbench/data/montgomery/fold_1_rgb.py
create mode 100644 src/ptbench/data/montgomery/fold_2.py
create mode 100644 src/ptbench/data/montgomery/fold_2_rgb.py
create mode 100644 src/ptbench/data/montgomery/fold_3.py
create mode 100644 src/ptbench/data/montgomery/fold_3_rgb.py
create mode 100644 src/ptbench/data/montgomery/fold_4.py
create mode 100644 src/ptbench/data/montgomery/fold_4_rgb.py
create mode 100644 src/ptbench/data/montgomery/fold_5.py
create mode 100644 src/ptbench/data/montgomery/fold_5_rgb.py
create mode 100644 src/ptbench/data/montgomery/fold_6.py
create mode 100644 src/ptbench/data/montgomery/fold_6_rgb.py
create mode 100644 src/ptbench/data/montgomery/fold_7.py
create mode 100644 src/ptbench/data/montgomery/fold_7_rgb.py
create mode 100644 src/ptbench/data/montgomery/fold_8.py
create mode 100644 src/ptbench/data/montgomery/fold_8_rgb.py
create mode 100644 src/ptbench/data/montgomery/fold_9.py
create mode 100644 src/ptbench/data/montgomery/fold_9_rgb.py
create mode 100644 src/ptbench/data/montgomery/rgb.py
diff --git a/pyproject.toml b/pyproject.toml
index 40c7a8fa..1becd87d 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -82,28 +82,28 @@ alexnet_pre = "ptbench.configs.models.alexnet_pretrained"
densenet = "ptbench.configs.models.densenet"
densenet_pre = "ptbench.configs.models.densenet_pretrained"
# montgomery dataset (and cross-validation folds)
-montgomery = "ptbench.configs.datasets.montgomery.default"
-montgomery_rgb = "ptbench.configs.datasets.montgomery.rgb"
-montgomery_f0 = "ptbench.configs.datasets.montgomery.fold_0"
-montgomery_f1 = "ptbench.configs.datasets.montgomery.fold_1"
-montgomery_f2 = "ptbench.configs.datasets.montgomery.fold_2"
-montgomery_f3 = "ptbench.configs.datasets.montgomery.fold_3"
-montgomery_f4 = "ptbench.configs.datasets.montgomery.fold_4"
-montgomery_f5 = "ptbench.configs.datasets.montgomery.fold_5"
-montgomery_f6 = "ptbench.configs.datasets.montgomery.fold_6"
-montgomery_f7 = "ptbench.configs.datasets.montgomery.fold_7"
-montgomery_f8 = "ptbench.configs.datasets.montgomery.fold_8"
-montgomery_f9 = "ptbench.configs.datasets.montgomery.fold_9"
-montgomery_f0_rgb = "ptbench.configs.datasets.montgomery.fold_0_rgb"
-montgomery_f1_rgb = "ptbench.configs.datasets.montgomery.fold_1_rgb"
-montgomery_f2_rgb = "ptbench.configs.datasets.montgomery.fold_2_rgb"
-montgomery_f3_rgb = "ptbench.configs.datasets.montgomery.fold_3_rgb"
-montgomery_f4_rgb = "ptbench.configs.datasets.montgomery.fold_4_rgb"
-montgomery_f5_rgb = "ptbench.configs.datasets.montgomery.fold_5_rgb"
-montgomery_f6_rgb = "ptbench.configs.datasets.montgomery.fold_6_rgb"
-montgomery_f7_rgb = "ptbench.configs.datasets.montgomery.fold_7_rgb"
-montgomery_f8_rgb = "ptbench.configs.datasets.montgomery.fold_8_rgb"
-montgomery_f9_rgb = "ptbench.configs.datasets.montgomery.fold_9_rgb"
+montgomery = "ptbench.data.montgomery.default"
+montgomery_rgb = "ptbench.data.montgomery.rgb"
+montgomery_f0 = "ptbench.data.montgomery.fold_0"
+montgomery_f1 = "ptbench.data.montgomery.montgomery.fold_1"
+montgomery_f2 = "ptbench.data.montgomery.fold_2"
+montgomery_f3 = "ptbench.data.montgomery.fold_3"
+montgomery_f4 = "ptbench.data.montgomery.fold_4"
+montgomery_f5 = "ptbench.data.montgomery.fold_5"
+montgomery_f6 = "ptbench.data.montgomery.fold_6"
+montgomery_f7 = "ptbench.data.montgomery.fold_7"
+montgomery_f8 = "ptbench.data.montgomery.fold_8"
+montgomery_f9 = "ptbench.data.montgomery.fold_9"
+montgomery_f0_rgb = "ptbench.data.montgomery.fold_0_rgb"
+montgomery_f1_rgb = "ptbench.data.montgomery.fold_1_rgb"
+montgomery_f2_rgb = "ptbench.data.montgomery.fold_2_rgb"
+montgomery_f3_rgb = "ptbench.data.montgomery.fold_3_rgb"
+montgomery_f4_rgb = "ptbench.data.montgomery.fold_4_rgb"
+montgomery_f5_rgb = "ptbench.data.montgomery.fold_5_rgb"
+montgomery_f6_rgb = "ptbench.data.montgomery.fold_6_rgb"
+montgomery_f7_rgb = "ptbench.data.montgomery.fold_7_rgb"
+montgomery_f8_rgb = "ptbench.data.montgomery.fold_8_rgb"
+montgomery_f9_rgb = "ptbench.data.montgomery.fold_9_rgb"
# extended montgomery dataset (with radiological signs)
montgomery_rs = "ptbench.configs.datasets.montgomery_RS.default"
montgomery_rs_f0 = "ptbench.configs.datasets.montgomery_RS.fold_0"
diff --git a/src/ptbench/configs/datasets/montgomery/__init__.py b/src/ptbench/configs/datasets/montgomery/__init__.py
deleted file mode 100644
index 86070a90..00000000
--- a/src/ptbench/configs/datasets/montgomery/__init__.py
+++ /dev/null
@@ -1,29 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-
-def _maker(protocol, resize_size=512, cc_size=512, RGB=False):
- from torchvision import transforms
-
- from ....data.montgomery import dataset as raw
- from ....data.transforms import ElasticDeformation, RemoveBlackBorders
- from .. import make_dataset as mk
-
- post_transforms = []
- if RGB:
- post_transforms = [
- transforms.Lambda(lambda x: x.convert("RGB")),
- transforms.ToTensor(),
- ]
-
- return mk(
- [raw.subsets(protocol)],
- [
- RemoveBlackBorders(),
- transforms.Resize(resize_size),
- transforms.CenterCrop(cc_size),
- ],
- [ElasticDeformation(p=0.8)],
- post_transforms,
- )
diff --git a/src/ptbench/configs/datasets/montgomery/default.py b/src/ptbench/configs/datasets/montgomery/default.py
deleted file mode 100644
index b00437e2..00000000
--- a/src/ptbench/configs/datasets/montgomery/default.py
+++ /dev/null
@@ -1,15 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Montgomery dataset for TB detection (default protocol)
-
-* Split reference: first 64% of TB and healthy CXR for "train" 16% for
-* "validation", 20% for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.montgomery` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("default")
diff --git a/src/ptbench/configs/datasets/montgomery/fold_0.py b/src/ptbench/configs/datasets/montgomery/fold_0.py
deleted file mode 100644
index a1d71237..00000000
--- a/src/ptbench/configs/datasets/montgomery/fold_0.py
+++ /dev/null
@@ -1,15 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Montgomery dataset for TB detection (cross validation fold 0)
-
-* Split reference: first 64% of TB and healthy CXR for "train" 16% for
-* "validation", 20% for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.montgomery` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_0")
diff --git a/src/ptbench/configs/datasets/montgomery/fold_0_rgb.py b/src/ptbench/configs/datasets/montgomery/fold_0_rgb.py
deleted file mode 100644
index 059dba2e..00000000
--- a/src/ptbench/configs/datasets/montgomery/fold_0_rgb.py
+++ /dev/null
@@ -1,15 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Montgomery dataset for TB detection (cross validation fold 0, RGB)
-
-* Split reference: first 64% of TB and healthy CXR for "train" 16% for
-* "validation", 20% for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.montgomery` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_0", RGB=True)
diff --git a/src/ptbench/configs/datasets/montgomery/fold_1.py b/src/ptbench/configs/datasets/montgomery/fold_1.py
deleted file mode 100644
index 4ee7db38..00000000
--- a/src/ptbench/configs/datasets/montgomery/fold_1.py
+++ /dev/null
@@ -1,15 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Montgomery dataset for TB detection (cross validation fold 1)
-
-* Split reference: first 64% of TB and healthy CXR for "train" 16% for
-* "validation", 20% for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.montgomery` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_1")
diff --git a/src/ptbench/configs/datasets/montgomery/fold_1_rgb.py b/src/ptbench/configs/datasets/montgomery/fold_1_rgb.py
deleted file mode 100644
index 9bc9214c..00000000
--- a/src/ptbench/configs/datasets/montgomery/fold_1_rgb.py
+++ /dev/null
@@ -1,15 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Montgomery dataset for TB detection (cross validation fold 1, RGB)
-
-* Split reference: first 64% of TB and healthy CXR for "train" 16% for
-* "validation", 20% for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.montgomery` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_1", RGB=True)
diff --git a/src/ptbench/configs/datasets/montgomery/fold_2.py b/src/ptbench/configs/datasets/montgomery/fold_2.py
deleted file mode 100644
index 6574b5de..00000000
--- a/src/ptbench/configs/datasets/montgomery/fold_2.py
+++ /dev/null
@@ -1,15 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Montgomery dataset for TB detection (cross validation fold 2)
-
-* Split reference: first 64% of TB and healthy CXR for "train" 16% for
-* "validation", 20% for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.montgomery` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_2")
diff --git a/src/ptbench/configs/datasets/montgomery/fold_2_rgb.py b/src/ptbench/configs/datasets/montgomery/fold_2_rgb.py
deleted file mode 100644
index 3423c925..00000000
--- a/src/ptbench/configs/datasets/montgomery/fold_2_rgb.py
+++ /dev/null
@@ -1,15 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Montgomery dataset for TB detection (cross validation fold 2, RGB)
-
-* Split reference: first 64% of TB and healthy CXR for "train" 16% for
-* "validation", 20% for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.montgomery` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_2", RGB=True)
diff --git a/src/ptbench/configs/datasets/montgomery/fold_3.py b/src/ptbench/configs/datasets/montgomery/fold_3.py
deleted file mode 100644
index 99d94f14..00000000
--- a/src/ptbench/configs/datasets/montgomery/fold_3.py
+++ /dev/null
@@ -1,15 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Montgomery dataset for TB detection (cross validation fold 3)
-
-* Split reference: first 64% of TB and healthy CXR for "train" 16% for
-* "validation", 20% for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.montgomery` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_3")
diff --git a/src/ptbench/configs/datasets/montgomery/fold_3_rgb.py b/src/ptbench/configs/datasets/montgomery/fold_3_rgb.py
deleted file mode 100644
index c1873c53..00000000
--- a/src/ptbench/configs/datasets/montgomery/fold_3_rgb.py
+++ /dev/null
@@ -1,15 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Montgomery dataset for TB detection (cross validation fold 3, RGB)
-
-* Split reference: first 64% of TB and healthy CXR for "train" 16% for
-* "validation", 20% for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.montgomery` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_3", RGB=True)
diff --git a/src/ptbench/configs/datasets/montgomery/fold_4.py b/src/ptbench/configs/datasets/montgomery/fold_4.py
deleted file mode 100644
index 42584bad..00000000
--- a/src/ptbench/configs/datasets/montgomery/fold_4.py
+++ /dev/null
@@ -1,15 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Montgomery dataset for TB detection (cross validation fold 4)
-
-* Split reference: first 64% of TB and healthy CXR for "train" 16% for
-* "validation", 20% for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.montgomery` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_4")
diff --git a/src/ptbench/configs/datasets/montgomery/fold_4_rgb.py b/src/ptbench/configs/datasets/montgomery/fold_4_rgb.py
deleted file mode 100644
index ed76962e..00000000
--- a/src/ptbench/configs/datasets/montgomery/fold_4_rgb.py
+++ /dev/null
@@ -1,15 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Montgomery dataset for TB detection (cross validation fold 4, RGB)
-
-* Split reference: first 64% of TB and healthy CXR for "train" 16% for
-* "validation", 20% for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.montgomery` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_4", RGB=True)
diff --git a/src/ptbench/configs/datasets/montgomery/fold_5.py b/src/ptbench/configs/datasets/montgomery/fold_5.py
deleted file mode 100644
index bb05a3f9..00000000
--- a/src/ptbench/configs/datasets/montgomery/fold_5.py
+++ /dev/null
@@ -1,15 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Montgomery dataset for TB detection (cross validation fold 5)
-
-* Split reference: first 64% of TB and healthy CXR for "train" 16% for
-* "validation", 20% for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.montgomery` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_5")
diff --git a/src/ptbench/configs/datasets/montgomery/fold_5_rgb.py b/src/ptbench/configs/datasets/montgomery/fold_5_rgb.py
deleted file mode 100644
index a65fd2c7..00000000
--- a/src/ptbench/configs/datasets/montgomery/fold_5_rgb.py
+++ /dev/null
@@ -1,15 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Montgomery dataset for TB detection (cross validation fold 5, RGB)
-
-* Split reference: first 64% of TB and healthy CXR for "train" 16% for
-* "validation", 20% for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.montgomery` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_5", RGB=True)
diff --git a/src/ptbench/configs/datasets/montgomery/fold_6.py b/src/ptbench/configs/datasets/montgomery/fold_6.py
deleted file mode 100644
index bbd97489..00000000
--- a/src/ptbench/configs/datasets/montgomery/fold_6.py
+++ /dev/null
@@ -1,15 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Montgomery dataset for TB detection (cross validation fold 6)
-
-* Split reference: first 64% of TB and healthy CXR for "train" 16% for
-* "validation", 20% for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.montgomery` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_6")
diff --git a/src/ptbench/configs/datasets/montgomery/fold_6_rgb.py b/src/ptbench/configs/datasets/montgomery/fold_6_rgb.py
deleted file mode 100644
index 89c7df33..00000000
--- a/src/ptbench/configs/datasets/montgomery/fold_6_rgb.py
+++ /dev/null
@@ -1,15 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Montgomery dataset for TB detection (cross validation fold 6, RGB)
-
-* Split reference: first 64% of TB and healthy CXR for "train" 16% for
-* "validation", 20% for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.montgomery` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_6", RGB=True)
diff --git a/src/ptbench/configs/datasets/montgomery/fold_7.py b/src/ptbench/configs/datasets/montgomery/fold_7.py
deleted file mode 100644
index 00ed7d68..00000000
--- a/src/ptbench/configs/datasets/montgomery/fold_7.py
+++ /dev/null
@@ -1,15 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Montgomery dataset for TB detection (cross validation fold 7)
-
-* Split reference: first 64% of TB and healthy CXR for "train" 16% for
-* "validation", 20% for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.montgomery` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_7")
diff --git a/src/ptbench/configs/datasets/montgomery/fold_7_rgb.py b/src/ptbench/configs/datasets/montgomery/fold_7_rgb.py
deleted file mode 100644
index 01a7c829..00000000
--- a/src/ptbench/configs/datasets/montgomery/fold_7_rgb.py
+++ /dev/null
@@ -1,16 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-
-"""Montgomery dataset for TB detection (cross validation fold 7, RGB)
-
-* Split reference: first 64% of TB and healthy CXR for "train" 16% for
-* "validation", 20% for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.montgomery` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_7", RGB=True)
diff --git a/src/ptbench/configs/datasets/montgomery/fold_8.py b/src/ptbench/configs/datasets/montgomery/fold_8.py
deleted file mode 100644
index 96f49d74..00000000
--- a/src/ptbench/configs/datasets/montgomery/fold_8.py
+++ /dev/null
@@ -1,15 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Montgomery dataset for TB detection (cross validation fold 8)
-
-* Split reference: first 64% of TB and healthy CXR for "train" 16% for
-* "validation", 20% for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.montgomery` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_8")
diff --git a/src/ptbench/configs/datasets/montgomery/fold_8_rgb.py b/src/ptbench/configs/datasets/montgomery/fold_8_rgb.py
deleted file mode 100644
index b9a2a556..00000000
--- a/src/ptbench/configs/datasets/montgomery/fold_8_rgb.py
+++ /dev/null
@@ -1,15 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Montgomery dataset for TB detection (cross validation fold 8, RGB)
-
-* Split reference: first 64% of TB and healthy CXR for "train" 16% for
-* "validation", 20% for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.montgomery` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_8", RGB=True)
diff --git a/src/ptbench/configs/datasets/montgomery/fold_9.py b/src/ptbench/configs/datasets/montgomery/fold_9.py
deleted file mode 100644
index 114392e4..00000000
--- a/src/ptbench/configs/datasets/montgomery/fold_9.py
+++ /dev/null
@@ -1,15 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Montgomery dataset for TB detection (cross validation fold 9)
-
-* Split reference: first 64% of TB and healthy CXR for "train" 16% for
-* "validation", 20% for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.montgomery` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_9")
diff --git a/src/ptbench/configs/datasets/montgomery/fold_9_rgb.py b/src/ptbench/configs/datasets/montgomery/fold_9_rgb.py
deleted file mode 100644
index df5340cb..00000000
--- a/src/ptbench/configs/datasets/montgomery/fold_9_rgb.py
+++ /dev/null
@@ -1,15 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Montgomery dataset for TB detection (cross validation fold 9, RGB)
-
-* Split reference: first 64% of TB and healthy CXR for "train" 16% for
-* "validation", 20% for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.montgomery` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_9", RGB=True)
diff --git a/src/ptbench/configs/datasets/montgomery/rgb.py b/src/ptbench/configs/datasets/montgomery/rgb.py
deleted file mode 100644
index 880ae2d1..00000000
--- a/src/ptbench/configs/datasets/montgomery/rgb.py
+++ /dev/null
@@ -1,15 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Montgomery dataset for TB detection (default protocol, converted in RGB)
-
-* Split reference: first 64% of TB and healthy CXR for "train" 16% for
-* "validation", 20% for "test"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.montgomery` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("default", RGB=True)
diff --git a/src/ptbench/data/montgomery/__init__.py b/src/ptbench/data/montgomery/__init__.py
index 38b50a26..65239cbf 100644
--- a/src/ptbench/data/montgomery/__init__.py
+++ b/src/ptbench/data/montgomery/__init__.py
@@ -22,6 +22,7 @@ import importlib.resources
import os
from ...utils.rc import load_rc
+from .. import make_dataset
from ..dataset import JSONDataset
from ..loader import load_pil_baw, make_delayed
@@ -55,9 +56,33 @@ def _loader(context, sample):
return make_delayed(sample, _raw_data_loader)
-dataset = JSONDataset(
+json_dataset = JSONDataset(
protocols=_protocols,
fieldnames=("data", "label"),
loader=_loader,
)
"""Montgomery dataset object."""
+
+
+def _maker(protocol, resize_size=512, cc_size=512, RGB=False):
+ from torchvision import transforms
+
+ from ..transforms import ElasticDeformation, 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/montgomery/default.py b/src/ptbench/data/montgomery/default.py
new file mode 100644
index 00000000..1f5c0809
--- /dev/null
+++ b/src/ptbench/data/montgomery/default.py
@@ -0,0 +1,47 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Montgomery dataset for TB detection (default protocol)
+
+* Split reference: first 64% of TB and healthy CXR for "train" 16% for
+* "validation", 20% for "test"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.montgomery` 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 DefaultModule(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("default")
+ (
+ self.train_dataset,
+ self.validation_dataset,
+ self.extra_validation_datasets,
+ self.predict_dataset,
+ ) = return_subsets(self.dataset)
+
+
+datamodule = DefaultModule
diff --git a/src/ptbench/data/montgomery/fold_0.py b/src/ptbench/data/montgomery/fold_0.py
new file mode 100644
index 00000000..c60791be
--- /dev/null
+++ b/src/ptbench/data/montgomery/fold_0.py
@@ -0,0 +1,47 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Montgomery dataset for TB detection (cross validation fold 0)
+
+* Split reference: first 64% of TB and healthy CXR for "train" 16% for
+* "validation", 20% for "test"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.montgomery` 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 DefaultModule(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)
+
+
+datamodule = DefaultModule
diff --git a/src/ptbench/data/montgomery/fold_0_rgb.py b/src/ptbench/data/montgomery/fold_0_rgb.py
new file mode 100644
index 00000000..8e8b0c89
--- /dev/null
+++ b/src/ptbench/data/montgomery/fold_0_rgb.py
@@ -0,0 +1,47 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Montgomery dataset for TB detection (cross validation fold 0, RGB)
+
+* Split reference: first 64% of TB and healthy CXR for "train" 16% for
+* "validation", 20% for "test"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.montgomery` 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 DefaultModule(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 = DefaultModule
diff --git a/src/ptbench/data/montgomery/fold_1.py b/src/ptbench/data/montgomery/fold_1.py
new file mode 100644
index 00000000..d6627e67
--- /dev/null
+++ b/src/ptbench/data/montgomery/fold_1.py
@@ -0,0 +1,47 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Montgomery dataset for TB detection (cross validation fold 1)
+
+* Split reference: first 64% of TB and healthy CXR for "train" 16% for
+* "validation", 20% for "test"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.montgomery` 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 DefaultModule(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)
+
+
+datamodule = DefaultModule
diff --git a/src/ptbench/data/montgomery/fold_1_rgb.py b/src/ptbench/data/montgomery/fold_1_rgb.py
new file mode 100644
index 00000000..bc47a322
--- /dev/null
+++ b/src/ptbench/data/montgomery/fold_1_rgb.py
@@ -0,0 +1,47 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Montgomery dataset for TB detection (cross validation fold 1, RGB)
+
+* Split reference: first 64% of TB and healthy CXR for "train" 16% for
+* "validation", 20% for "test"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.montgomery` 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 DefaultModule(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 = DefaultModule
diff --git a/src/ptbench/data/montgomery/fold_2.py b/src/ptbench/data/montgomery/fold_2.py
new file mode 100644
index 00000000..8c5f4a66
--- /dev/null
+++ b/src/ptbench/data/montgomery/fold_2.py
@@ -0,0 +1,47 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Montgomery dataset for TB detection (cross validation fold 2)
+
+* Split reference: first 64% of TB and healthy CXR for "train" 16% for
+* "validation", 20% for "test"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.montgomery` 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 DefaultModule(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)
+
+
+datamodule = DefaultModule
diff --git a/src/ptbench/data/montgomery/fold_2_rgb.py b/src/ptbench/data/montgomery/fold_2_rgb.py
new file mode 100644
index 00000000..b81a877b
--- /dev/null
+++ b/src/ptbench/data/montgomery/fold_2_rgb.py
@@ -0,0 +1,47 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Montgomery dataset for TB detection (cross validation fold 2, RGB)
+
+* Split reference: first 64% of TB and healthy CXR for "train" 16% for
+* "validation", 20% for "test"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.montgomery` 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 DefaultModule(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 = DefaultModule
diff --git a/src/ptbench/data/montgomery/fold_3.py b/src/ptbench/data/montgomery/fold_3.py
new file mode 100644
index 00000000..8e685d7e
--- /dev/null
+++ b/src/ptbench/data/montgomery/fold_3.py
@@ -0,0 +1,47 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Montgomery dataset for TB detection (cross validation fold 3)
+
+* Split reference: first 64% of TB and healthy CXR for "train" 16% for
+* "validation", 20% for "test"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.montgomery` 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 DefaultModule(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)
+
+
+datamodule = DefaultModule
diff --git a/src/ptbench/data/montgomery/fold_3_rgb.py b/src/ptbench/data/montgomery/fold_3_rgb.py
new file mode 100644
index 00000000..7b600371
--- /dev/null
+++ b/src/ptbench/data/montgomery/fold_3_rgb.py
@@ -0,0 +1,47 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Montgomery dataset for TB detection (cross validation fold 3, RGB)
+
+* Split reference: first 64% of TB and healthy CXR for "train" 16% for
+* "validation", 20% for "test"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.montgomery` 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 DefaultModule(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 = DefaultModule
diff --git a/src/ptbench/data/montgomery/fold_4.py b/src/ptbench/data/montgomery/fold_4.py
new file mode 100644
index 00000000..9459cb93
--- /dev/null
+++ b/src/ptbench/data/montgomery/fold_4.py
@@ -0,0 +1,47 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Montgomery dataset for TB detection (cross validation fold 4)
+
+* Split reference: first 64% of TB and healthy CXR for "train" 16% for
+* "validation", 20% for "test"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.montgomery` 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 DefaultModule(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)
+
+
+datamodule = DefaultModule
diff --git a/src/ptbench/data/montgomery/fold_4_rgb.py b/src/ptbench/data/montgomery/fold_4_rgb.py
new file mode 100644
index 00000000..3eb136f6
--- /dev/null
+++ b/src/ptbench/data/montgomery/fold_4_rgb.py
@@ -0,0 +1,47 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Montgomery dataset for TB detection (cross validation fold 4, RGB)
+
+* Split reference: first 64% of TB and healthy CXR for "train" 16% for
+* "validation", 20% for "test"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.montgomery` 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 DefaultModule(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 = DefaultModule
diff --git a/src/ptbench/data/montgomery/fold_5.py b/src/ptbench/data/montgomery/fold_5.py
new file mode 100644
index 00000000..147690f6
--- /dev/null
+++ b/src/ptbench/data/montgomery/fold_5.py
@@ -0,0 +1,47 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Montgomery dataset for TB detection (cross validation fold 5)
+
+* Split reference: first 64% of TB and healthy CXR for "train" 16% for
+* "validation", 20% for "test"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.montgomery` 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 DefaultModule(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)
+
+
+datamodule = DefaultModule
diff --git a/src/ptbench/data/montgomery/fold_5_rgb.py b/src/ptbench/data/montgomery/fold_5_rgb.py
new file mode 100644
index 00000000..3e7cb73f
--- /dev/null
+++ b/src/ptbench/data/montgomery/fold_5_rgb.py
@@ -0,0 +1,47 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Montgomery dataset for TB detection (cross validation fold 5, RGB)
+
+* Split reference: first 64% of TB and healthy CXR for "train" 16% for
+* "validation", 20% for "test"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.montgomery` 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 DefaultModule(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 = DefaultModule
diff --git a/src/ptbench/data/montgomery/fold_6.py b/src/ptbench/data/montgomery/fold_6.py
new file mode 100644
index 00000000..69f24390
--- /dev/null
+++ b/src/ptbench/data/montgomery/fold_6.py
@@ -0,0 +1,47 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Montgomery dataset for TB detection (cross validation fold 6)
+
+* Split reference: first 64% of TB and healthy CXR for "train" 16% for
+* "validation", 20% for "test"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.montgomery` 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 DefaultModule(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)
+
+
+datamodule = DefaultModule
diff --git a/src/ptbench/data/montgomery/fold_6_rgb.py b/src/ptbench/data/montgomery/fold_6_rgb.py
new file mode 100644
index 00000000..ff3a8cdb
--- /dev/null
+++ b/src/ptbench/data/montgomery/fold_6_rgb.py
@@ -0,0 +1,47 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Montgomery dataset for TB detection (cross validation fold 6, RGB)
+
+* Split reference: first 64% of TB and healthy CXR for "train" 16% for
+* "validation", 20% for "test"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.montgomery` 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 DefaultModule(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 = DefaultModule
diff --git a/src/ptbench/data/montgomery/fold_7.py b/src/ptbench/data/montgomery/fold_7.py
new file mode 100644
index 00000000..20ba9d3a
--- /dev/null
+++ b/src/ptbench/data/montgomery/fold_7.py
@@ -0,0 +1,47 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Montgomery dataset for TB detection (cross validation fold 7)
+
+* Split reference: first 64% of TB and healthy CXR for "train" 16% for
+* "validation", 20% for "test"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.montgomery` 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 DefaultModule(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)
+
+
+datamodule = DefaultModule
diff --git a/src/ptbench/data/montgomery/fold_7_rgb.py b/src/ptbench/data/montgomery/fold_7_rgb.py
new file mode 100644
index 00000000..05664b06
--- /dev/null
+++ b/src/ptbench/data/montgomery/fold_7_rgb.py
@@ -0,0 +1,47 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Montgomery dataset for TB detection (cross validation fold 7, RGB)
+
+* Split reference: first 64% of TB and healthy CXR for "train" 16% for
+* "validation", 20% for "test"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.montgomery` 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 DefaultModule(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 = DefaultModule
diff --git a/src/ptbench/data/montgomery/fold_8.py b/src/ptbench/data/montgomery/fold_8.py
new file mode 100644
index 00000000..e92ff959
--- /dev/null
+++ b/src/ptbench/data/montgomery/fold_8.py
@@ -0,0 +1,47 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Montgomery dataset for TB detection (cross validation fold 8)
+
+* Split reference: first 64% of TB and healthy CXR for "train" 16% for
+* "validation", 20% for "test"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.montgomery` 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 DefaultModule(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)
+
+
+datamodule = DefaultModule
diff --git a/src/ptbench/data/montgomery/fold_8_rgb.py b/src/ptbench/data/montgomery/fold_8_rgb.py
new file mode 100644
index 00000000..b7d59359
--- /dev/null
+++ b/src/ptbench/data/montgomery/fold_8_rgb.py
@@ -0,0 +1,47 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Montgomery dataset for TB detection (cross validation fold 8, RGB)
+
+* Split reference: first 64% of TB and healthy CXR for "train" 16% for
+* "validation", 20% for "test"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.montgomery` 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 DefaultModule(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 = DefaultModule
diff --git a/src/ptbench/data/montgomery/fold_9.py b/src/ptbench/data/montgomery/fold_9.py
new file mode 100644
index 00000000..81bbf72e
--- /dev/null
+++ b/src/ptbench/data/montgomery/fold_9.py
@@ -0,0 +1,47 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Montgomery dataset for TB detection (cross validation fold 9)
+
+* Split reference: first 64% of TB and healthy CXR for "train" 16% for
+* "validation", 20% for "test"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.montgomery` 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 DefaultModule(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)
+
+
+datamodule = DefaultModule
diff --git a/src/ptbench/data/montgomery/fold_9_rgb.py b/src/ptbench/data/montgomery/fold_9_rgb.py
new file mode 100644
index 00000000..e961e08f
--- /dev/null
+++ b/src/ptbench/data/montgomery/fold_9_rgb.py
@@ -0,0 +1,47 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Montgomery dataset for TB detection (cross validation fold 9, RGB)
+
+* Split reference: first 64% of TB and healthy CXR for "train" 16% for
+* "validation", 20% for "test"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.montgomery` 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 DefaultModule(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 = DefaultModule
diff --git a/src/ptbench/data/montgomery/rgb.py b/src/ptbench/data/montgomery/rgb.py
new file mode 100644
index 00000000..c1621266
--- /dev/null
+++ b/src/ptbench/data/montgomery/rgb.py
@@ -0,0 +1,47 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Montgomery dataset for TB detection (default protocol, converted in RGB)
+
+* Split reference: first 64% of TB and healthy CXR for "train" 16% for
+* "validation", 20% for "test"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.montgomery` 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 DefaultModule(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("default", RGB=True)
+ (
+ self.train_dataset,
+ self.validation_dataset,
+ self.extra_validation_datasets,
+ self.predict_dataset,
+ ) = return_subsets(self.dataset)
+
+
+datamodule = DefaultModule
--
GitLab