diff --git a/pyproject.toml b/pyproject.toml
index 27ef33fc79192665641364a77e2fbcdd613a5398..1aee2d658802aae55f37e7307f91cbb2448d5380 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -222,28 +222,28 @@ tbx11k_simplified_rs_f7 = "ptbench.configs.datasets.tbx11k_simplified_RS.fold_7"
tbx11k_simplified_rs_f8 = "ptbench.configs.datasets.tbx11k_simplified_RS.fold_8"
tbx11k_simplified_rs_f9 = "ptbench.configs.datasets.tbx11k_simplified_RS.fold_9"
# TBX11K simplified dataset split 2 (and cross-validation folds)
-tbx11k_simplified_v2 = "ptbench.configs.datasets.tbx11k_simplified_v2.default"
-tbx11k_simplified_v2_rgb = "ptbench.configs.datasets.tbx11k_simplified_v2.rgb"
-tbx11k_simplified_v2_f0 = "ptbench.configs.datasets.tbx11k_simplified_v2.fold_0"
-tbx11k_simplified_v2_f1 = "ptbench.configs.datasets.tbx11k_simplified_v2.fold_1"
-tbx11k_simplified_v2_f2 = "ptbench.configs.datasets.tbx11k_simplified_v2.fold_2"
-tbx11k_simplified_v2_f3 = "ptbench.configs.datasets.tbx11k_simplified_v2.fold_3"
-tbx11k_simplified_v2_f4 = "ptbench.configs.datasets.tbx11k_simplified_v2.fold_4"
-tbx11k_simplified_v2_f5 = "ptbench.configs.datasets.tbx11k_simplified_v2.fold_5"
-tbx11k_simplified_v2_f6 = "ptbench.configs.datasets.tbx11k_simplified_v2.fold_6"
-tbx11k_simplified_v2_f7 = "ptbench.configs.datasets.tbx11k_simplified_v2.fold_7"
-tbx11k_simplified_v2_f8 = "ptbench.configs.datasets.tbx11k_simplified_v2.fold_8"
-tbx11k_simplified_v2_f9 = "ptbench.configs.datasets.tbx11k_simplified_v2.fold_9"
-tbx11k_simplified_v2_f0_rgb = "ptbench.configs.datasets.tbx11k_simplified_v2.fold_0_rgb"
-tbx11k_simplified_v2_f1_rgb = "ptbench.configs.datasets.tbx11k_simplified_v2.fold_1_rgb"
-tbx11k_simplified_v2_f2_rgb = "ptbench.configs.datasets.tbx11k_simplified_v2.fold_2_rgb"
-tbx11k_simplified_v2_f3_rgb = "ptbench.configs.datasets.tbx11k_simplified_v2.fold_3_rgb"
-tbx11k_simplified_v2_f4_rgb = "ptbench.configs.datasets.tbx11k_simplified_v2.fold_4_rgb"
-tbx11k_simplified_v2_f5_rgb = "ptbench.configs.datasets.tbx11k_simplified_v2.fold_5_rgb"
-tbx11k_simplified_v2_f6_rgb = "ptbench.configs.datasets.tbx11k_simplified_v2.fold_6_rgb"
-tbx11k_simplified_v2_f7_rgb = "ptbench.configs.datasets.tbx11k_simplified_v2.fold_7_rgb"
-tbx11k_simplified_v2_f8_rgb = "ptbench.configs.datasets.tbx11k_simplified_v2.fold_8_rgb"
-tbx11k_simplified_v2_f9_rgb = "ptbench.configs.datasets.tbx11k_simplified_v2.fold_9_rgb"
+tbx11k_simplified_v2 = "ptbench.data.tbx11k_simplified_v2.default"
+tbx11k_simplified_v2_rgb = "ptbench.data.tbx11k_simplified_v2.rgb"
+tbx11k_simplified_v2_f0 = "ptbench.data.tbx11k_simplified_v2.fold_0"
+tbx11k_simplified_v2_f1 = "ptbench.data.tbx11k_simplified_v2.fold_1"
+tbx11k_simplified_v2_f2 = "ptbench.data.tbx11k_simplified_v2.fold_2"
+tbx11k_simplified_v2_f3 = "ptbench.data.tbx11k_simplified_v2.fold_3"
+tbx11k_simplified_v2_f4 = "ptbench.data.tbx11k_simplified_v2.fold_4"
+tbx11k_simplified_v2_f5 = "ptbench.data.tbx11k_simplified_v2.fold_5"
+tbx11k_simplified_v2_f6 = "ptbench.data.tbx11k_simplified_v2.fold_6"
+tbx11k_simplified_v2_f7 = "ptbench.data.tbx11k_simplified_v2.fold_7"
+tbx11k_simplified_v2_f8 = "ptbench.data.tbx11k_simplified_v2.fold_8"
+tbx11k_simplified_v2_f9 = "ptbench.data.tbx11k_simplified_v2.fold_9"
+tbx11k_simplified_v2_f0_rgb = "ptbench.data.tbx11k_simplified_v2.fold_0_rgb"
+tbx11k_simplified_v2_f1_rgb = "ptbench.data.tbx11k_simplified_v2.fold_1_rgb"
+tbx11k_simplified_v2_f2_rgb = "ptbench.data.tbx11k_simplified_v2.fold_2_rgb"
+tbx11k_simplified_v2_f3_rgb = "ptbench.data.tbx11k_simplified_v2.fold_3_rgb"
+tbx11k_simplified_v2_f4_rgb = "ptbench.data.tbx11k_simplified_v2.fold_4_rgb"
+tbx11k_simplified_v2_f5_rgb = "ptbench.data.tbx11k_simplified_v2.fold_5_rgb"
+tbx11k_simplified_v2_f6_rgb = "ptbench.data.tbx11k_simplified_v2.fold_6_rgb"
+tbx11k_simplified_v2_f7_rgb = "ptbench.data.tbx11k_simplified_v2.fold_7_rgb"
+tbx11k_simplified_v2_f8_rgb = "ptbench.data.tbx11k_simplified_v2.fold_8_rgb"
+tbx11k_simplified_v2_f9_rgb = "ptbench.data.tbx11k_simplified_v2.fold_9_rgb"
# extended TBX11K simplified dataset split 2 (with radiological signs)
tbx11k_simplified_v2_rs = "ptbench.configs.datasets.tbx11k_simplified_v2_RS.default"
tbx11k_simplified_v2_rs_f0 = "ptbench.configs.datasets.tbx11k_simplified_v2_RS.fold_0"
diff --git a/src/ptbench/configs/datasets/tbx11k_simplified_v2/__init__.py b/src/ptbench/configs/datasets/tbx11k_simplified_v2/__init__.py
deleted file mode 100644
index a1e6f50ca64a471417fef6ba47253cce9af9b12f..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/tbx11k_simplified_v2/__init__.py
+++ /dev/null
@@ -1,25 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-
-def _maker(protocol, RGB=False):
- from torchvision import transforms
-
- from ....data.tbx11k_simplified_v2 import dataset as raw
- from ....data.transforms import ElasticDeformation
- 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)],
- [],
- [ElasticDeformation(p=0.8)],
- post_transforms,
- )
diff --git a/src/ptbench/configs/datasets/tbx11k_simplified_v2/default.py b/src/ptbench/configs/datasets/tbx11k_simplified_v2/default.py
deleted file mode 100644
index f969c8554d7b67f35b2aea6c08a09d2aeb7340bf..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/tbx11k_simplified_v2/default.py
+++ /dev/null
@@ -1,17 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""TBX11k simplified dataset for TB detection (default protocol)
-
-* Split reference: first 62.6% of CXR for "train", 16% for "validation",
-* 21.4% for "test"
-* This split consists of non-TB and active TB samples
-* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbx11k_simplified_v2` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("default")
diff --git a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_0.py b/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_0.py
deleted file mode 100644
index 1805c0005fbad4f93dbb8f2b4225959cb0882832..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_0.py
+++ /dev/null
@@ -1,17 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""TBX11k simplified dataset for TB detection (cross validation fold 0)
-
-* Split reference: first 62.6% of CXR for "train", 16% for "validation",
-* 21.4% for "test"
-* This split consists of non-TB and active TB samples
-* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbx11k_simplified_v2` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_0")
diff --git a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_0_rgb.py b/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_0_rgb.py
deleted file mode 100644
index b232652f8d009bd5ec8b80108201b216d3ada961..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_0_rgb.py
+++ /dev/null
@@ -1,17 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""TBX11k simplified dataset for TB detection (cross validation fold 0, RGB)
-
-* Split reference: first 62.6% of CXR for "train", 16% for "validation",
-* 21.4% for "test"
-* This split consists of non-TB and active TB samples
-* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbx11k_simplified_v2` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_0", RGB=True)
diff --git a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_1.py b/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_1.py
deleted file mode 100644
index 8d8074a1281c4b528527e02a824892239d558c1a..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_1.py
+++ /dev/null
@@ -1,17 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""TBX11k simplified dataset for TB detection (cross validation fold 1)
-
-* Split reference: first 62.6% of CXR for "train", 16% for "validation",
-* 21.4% for "test"
-* This split consists of non-TB and active TB samples
-* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbx11k_simplified_v2` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_1")
diff --git a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_1_rgb.py b/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_1_rgb.py
deleted file mode 100644
index ec0421059f94bffb2fa769ad5990f31636e75a47..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_1_rgb.py
+++ /dev/null
@@ -1,17 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""TBX11k simplified dataset for TB detection (cross validation fold 1, RGB)
-
-* Split reference: first 62.6% of CXR for "train", 16% for "validation",
-* 21.4% for "test"
-* This split consists of non-TB and active TB samples
-* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbx11k_simplified_v2` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_1", RGB=True)
diff --git a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_2.py b/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_2.py
deleted file mode 100644
index 1962b0370b9a3c24fbe1cb81119760f5239d4f83..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_2.py
+++ /dev/null
@@ -1,17 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""TBX11k simplified dataset for TB detection (cross validation fold 2)
-
-* Split reference: first 62.6% of CXR for "train", 16% for "validation",
-* 21.4% for "test"
-* This split consists of non-TB and active TB samples
-* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbx11k_simplified_v2` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_2")
diff --git a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_2_rgb.py b/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_2_rgb.py
deleted file mode 100644
index 660073a4c13ae62dba576cffb61953e063bdc560..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_2_rgb.py
+++ /dev/null
@@ -1,17 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""TBX11k simplified dataset for TB detection (cross validation fold 2, RGB)
-
-* Split reference: first 62.6% of CXR for "train", 16% for "validation",
-* 21.4% for "test"
-* This split consists of non-TB and active TB samples
-* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbx11k_simplified_v2` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_2", RGB=True)
diff --git a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_3.py b/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_3.py
deleted file mode 100644
index 9872ef5831ffb43a61afce0764a91915aae603d6..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_3.py
+++ /dev/null
@@ -1,17 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""TBX11k simplified dataset for TB detection (cross validation fold 3)
-
-* Split reference: first 62.6% of CXR for "train", 16% for "validation",
-* 21.4% for "test"
-* This split consists of non-TB and active TB samples
-* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbx11k_simplified_v2` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_3")
diff --git a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_3_rgb.py b/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_3_rgb.py
deleted file mode 100644
index bb54f1d206ea1893f1c6f4bac516e7d6a00d3a7d..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_3_rgb.py
+++ /dev/null
@@ -1,17 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""TBX11k simplified dataset for TB detection (cross validation fold 3, RGB)
-
-* Split reference: first 62.6% of CXR for "train", 16% for "validation",
-* 21.4% for "test"
-* This split consists of non-TB and active TB samples
-* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbx11k_simplified_v2` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_3", RGB=True)
diff --git a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_4.py b/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_4.py
deleted file mode 100644
index 7cde75d8faf0d9834bdd9c4bbc0715b76cc5c530..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_4.py
+++ /dev/null
@@ -1,17 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""TBX11k simplified dataset for TB detection (cross validation fold 4)
-
-* Split reference: first 62.6% of CXR for "train", 16% for "validation",
-* 21.4% for "test"
-* This split consists of non-TB and active TB samples
-* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbx11k_simplified_v2` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_4")
diff --git a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_4_rgb.py b/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_4_rgb.py
deleted file mode 100644
index d5bf8d8255148537ee2e432e4102949ee2f81d15..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_4_rgb.py
+++ /dev/null
@@ -1,17 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""TBX11k simplified dataset for TB detection (cross validation fold 4, RGB)
-
-* Split reference: first 62.6% of CXR for "train", 16% for "validation",
-* 21.4% for "test"
-* This split consists of non-TB and active TB samples
-* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbx11k_simplified_v2` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_4", RGB=True)
diff --git a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_5.py b/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_5.py
deleted file mode 100644
index d9ca35bf50da984d837d1e4609cb5178a6c6d39c..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_5.py
+++ /dev/null
@@ -1,17 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""TBX11k simplified dataset for TB detection (cross validation fold 5)
-
-* Split reference: first 62.6% of CXR for "train", 16% for "validation",
-* 21.4% for "test"
-* This split consists of non-TB and active TB samples
-* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbx11k_simplified_v2` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_5")
diff --git a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_5_rgb.py b/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_5_rgb.py
deleted file mode 100644
index 983326689b110a2ccb0d78fcfa61cf6a950d2d36..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_5_rgb.py
+++ /dev/null
@@ -1,17 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""TBX11k simplified dataset for TB detection (cross validation fold 5, RGB)
-
-* Split reference: first 62.6% of CXR for "train", 16% for "validation",
-* 21.4% for "test"
-* This split consists of non-TB and active TB samples
-* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbx11k_simplified_v2` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_5", RGB=True)
diff --git a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_6.py b/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_6.py
deleted file mode 100644
index c8abb0658bf78ad2cede7f1c4329735595284138..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_6.py
+++ /dev/null
@@ -1,17 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""TBX11k simplified dataset for TB detection (cross validation fold 6)
-
-* Split reference: first 62.6% of CXR for "train", 16% for "validation",
-* 21.4% for "test"
-* This split consists of non-TB and active TB samples
-* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbx11k_simplified_v2` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_6")
diff --git a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_6_rgb.py b/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_6_rgb.py
deleted file mode 100644
index dabdb67452b3cf30e31ab2a70cc08f30d2afb003..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_6_rgb.py
+++ /dev/null
@@ -1,17 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""TBX11k simplified dataset for TB detection (cross validation fold 6, RGB)
-
-* Split reference: first 62.6% of CXR for "train", 16% for "validation",
-* 21.4% for "test"
-* This split consists of non-TB and active TB samples
-* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbx11k_simplified_v2` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_6", RGB=True)
diff --git a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_7.py b/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_7.py
deleted file mode 100644
index 67864aa6ccdc3448177f7139879b0eb05f6cd10d..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_7.py
+++ /dev/null
@@ -1,17 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""TBX11k simplified dataset for TB detection (cross validation fold 7)
-
-* Split reference: first 62.6% of CXR for "train", 16% for "validation",
-* 21.4% for "test"
-* This split consists of non-TB and active TB samples
-* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbx11k_simplified_v2` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_7")
diff --git a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_7_rgb.py b/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_7_rgb.py
deleted file mode 100644
index d37bda6e40f650bc313c90159838857f4140d395..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_7_rgb.py
+++ /dev/null
@@ -1,17 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""TBX11k simplified dataset for TB detection (cross validation fold 7, RGB)
-
-* Split reference: first 62.6% of CXR for "train", 16% for "validation",
-* 21.4% for "test"
-* This split consists of non-TB and active TB samples
-* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbx11k_simplified_v2` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_7", RGB=True)
diff --git a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_8.py b/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_8.py
deleted file mode 100644
index 2ac58e16e9e9ebbb9dbc73432605c10c7ad53371..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_8.py
+++ /dev/null
@@ -1,17 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""TBX11k simplified dataset for TB detection (cross validation fold 8)
-
-* Split reference: first 62.6% of CXR for "train", 16% for "validation",
-* 21.4% for "test"
-* This split consists of non-TB and active TB samples
-* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbx11k_simplified_v2` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_8")
diff --git a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_8_rgb.py b/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_8_rgb.py
deleted file mode 100644
index f6adcd4300305ab0b153041444f17b1eef2a7de0..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_8_rgb.py
+++ /dev/null
@@ -1,17 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""TBX11k simplified dataset for TB detection (cross validation fold 8, RGB)
-
-* Split reference: first 62.6% of CXR for "train", 16% for "validation",
-* 21.4% for "test"
-* This split consists of non-TB and active TB samples
-* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbx11k_simplified_v2` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_8", RGB=True)
diff --git a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_9.py b/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_9.py
deleted file mode 100644
index 1034cf8176641e8afeb27fc2e00acea306411265..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_9.py
+++ /dev/null
@@ -1,17 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""TBX11k simplified dataset for TB detection (cross validation fold 9)
-
-* Split reference: first 62.6% of CXR for "train", 16% for "validation",
-* 21.4% for "test"
-* This split consists of non-TB and active TB samples
-* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbx11k_simplified_v2` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_9")
diff --git a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_9_rgb.py b/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_9_rgb.py
deleted file mode 100644
index 8fe3f1dbf75faa2537aabe297b41b82d4c9b4c3b..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/tbx11k_simplified_v2/fold_9_rgb.py
+++ /dev/null
@@ -1,17 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""TBX11k simplified dataset for TB detection (cross validation fold 9, RGB)
-
-* Split reference: first 62.6% of CXR for "train", 16% for "validation",
-* 21.4% for "test"
-* This split consists of non-TB and active TB samples
-* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbx11k_simplified_v2` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("fold_9", RGB=True)
diff --git a/src/ptbench/configs/datasets/tbx11k_simplified_v2/rgb.py b/src/ptbench/configs/datasets/tbx11k_simplified_v2/rgb.py
deleted file mode 100644
index c327dad9578765a4542aa1a41ce16d36d6d49d6a..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/tbx11k_simplified_v2/rgb.py
+++ /dev/null
@@ -1,18 +0,0 @@
-# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""TBX11k simplified dataset for TB detection (default protocol, converted in
-RGB)
-
-* Split reference: first 62.6% of CXR for "train", 16% for "validation",
-* 21.4% for "test"
-* This split consists of non-TB and active TB samples
-* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
-* This configuration resolution: 512 x 512 (default)
-* See :py:mod:`ptbench.data.tbx11k_simplified_v2` for dataset details
-"""
-
-from . import _maker
-
-dataset = _maker("default", RGB=True)
diff --git a/src/ptbench/data/tbx11k_simplified_v2/__init__.py b/src/ptbench/data/tbx11k_simplified_v2/__init__.py
index 7cf676b00d4570c8b2f6b42724f55eeb0eab2fc4..d6075e85d366510d8670af362a2b685167e3c8de 100644
--- a/src/ptbench/data/tbx11k_simplified_v2/__init__.py
+++ b/src/ptbench/data/tbx11k_simplified_v2/__init__.py
@@ -76,15 +76,36 @@ def _loader_bbox(context, sample):
return make_delayed_bbox(sample, _raw_data_loader_bbox)
-dataset = JSONDataset(
+json_dataset = JSONDataset(
protocols=_protocols,
fieldnames=("data", "label"),
loader=_loader,
)
-dataset_with_bboxes = JSONDataset(
+json_dataset_with_bboxes = JSONDataset(
protocols=_protocols,
fieldnames=("data", "label", "bboxes"),
loader=_loader_bbox,
)
"""TBX11K simplified dataset object."""
+
+
+def _maker(protocol, RGB=False):
+ from torchvision import transforms
+
+ from .. import make_dataset
+ from ..transforms import ElasticDeformation
+
+ post_transforms = []
+ if RGB:
+ post_transforms = [
+ transforms.Lambda(lambda x: x.convert("RGB")),
+ transforms.ToTensor(),
+ ]
+
+ return make_dataset(
+ [json_dataset.subsets(protocol)],
+ [],
+ [ElasticDeformation(p=0.8)],
+ post_transforms,
+ )
diff --git a/src/ptbench/data/tbx11k_simplified_v2/default.py b/src/ptbench/data/tbx11k_simplified_v2/default.py
new file mode 100644
index 0000000000000000000000000000000000000000..17b738c59be7a4a482ab479673da3a06e68240ea
--- /dev/null
+++ b/src/ptbench/data/tbx11k_simplified_v2/default.py
@@ -0,0 +1,49 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""TBX11k simplified dataset for TB detection (default protocol)
+
+* Split reference: first 62.6% of CXR for "train", 16% for "validation",
+* 21.4% for "test"
+* This split consists of non-TB and active TB samples
+* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.tbx11k_simplified_v2` 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/tbx11k_simplified_v2/fold_0.py b/src/ptbench/data/tbx11k_simplified_v2/fold_0.py
new file mode 100644
index 0000000000000000000000000000000000000000..4751d17ec8266170aca6519cf8e917e90305590f
--- /dev/null
+++ b/src/ptbench/data/tbx11k_simplified_v2/fold_0.py
@@ -0,0 +1,49 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""TBX11k simplified dataset for TB detection (cross validation fold 0)
+
+* Split reference: first 62.6% of CXR for "train", 16% for "validation",
+* 21.4% for "test"
+* This split consists of non-TB and active TB samples
+* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.tbx11k_simplified_v2` 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")
+ (
+ 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/tbx11k_simplified_v2/fold_0_rgb.py b/src/ptbench/data/tbx11k_simplified_v2/fold_0_rgb.py
new file mode 100644
index 0000000000000000000000000000000000000000..1ba0e2ac26cc63843b90d32b011a0eefc6a6d364
--- /dev/null
+++ b/src/ptbench/data/tbx11k_simplified_v2/fold_0_rgb.py
@@ -0,0 +1,49 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""TBX11k simplified dataset for TB detection (cross validation fold 0, RGB)
+
+* Split reference: first 62.6% of CXR for "train", 16% for "validation",
+* 21.4% for "test"
+* This split consists of non-TB and active TB samples
+* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.tbx11k_simplified_v2` 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/tbx11k_simplified_v2/fold_1.py b/src/ptbench/data/tbx11k_simplified_v2/fold_1.py
new file mode 100644
index 0000000000000000000000000000000000000000..c70e2d194c26cfd9e44fb68bb1423485bf27d0b3
--- /dev/null
+++ b/src/ptbench/data/tbx11k_simplified_v2/fold_1.py
@@ -0,0 +1,49 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""TBX11k simplified dataset for TB detection (cross validation fold 1)
+
+* Split reference: first 62.6% of CXR for "train", 16% for "validation",
+* 21.4% for "test"
+* This split consists of non-TB and active TB samples
+* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.tbx11k_simplified_v2` 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")
+ (
+ 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/tbx11k_simplified_v2/fold_1_rgb.py b/src/ptbench/data/tbx11k_simplified_v2/fold_1_rgb.py
new file mode 100644
index 0000000000000000000000000000000000000000..c543f62ef12c9bff50b1e457ea0aef2fde40f3f3
--- /dev/null
+++ b/src/ptbench/data/tbx11k_simplified_v2/fold_1_rgb.py
@@ -0,0 +1,49 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""TBX11k simplified dataset for TB detection (cross validation fold 1, RGB)
+
+* Split reference: first 62.6% of CXR for "train", 16% for "validation",
+* 21.4% for "test"
+* This split consists of non-TB and active TB samples
+* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.tbx11k_simplified_v2` 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/tbx11k_simplified_v2/fold_2.py b/src/ptbench/data/tbx11k_simplified_v2/fold_2.py
new file mode 100644
index 0000000000000000000000000000000000000000..5af836ee4090df17b5e3cf9093f13bfc8bae433f
--- /dev/null
+++ b/src/ptbench/data/tbx11k_simplified_v2/fold_2.py
@@ -0,0 +1,49 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""TBX11k simplified dataset for TB detection (cross validation fold 2)
+
+* Split reference: first 62.6% of CXR for "train", 16% for "validation",
+* 21.4% for "test"
+* This split consists of non-TB and active TB samples
+* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.tbx11k_simplified_v2` 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")
+ (
+ 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/tbx11k_simplified_v2/fold_2_rgb.py b/src/ptbench/data/tbx11k_simplified_v2/fold_2_rgb.py
new file mode 100644
index 0000000000000000000000000000000000000000..39ba5a97cb95826b706e36cc356995ba5ecd95a4
--- /dev/null
+++ b/src/ptbench/data/tbx11k_simplified_v2/fold_2_rgb.py
@@ -0,0 +1,49 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""TBX11k simplified dataset for TB detection (cross validation fold 2, RGB)
+
+* Split reference: first 62.6% of CXR for "train", 16% for "validation",
+* 21.4% for "test"
+* This split consists of non-TB and active TB samples
+* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.tbx11k_simplified_v2` 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/tbx11k_simplified_v2/fold_3.py b/src/ptbench/data/tbx11k_simplified_v2/fold_3.py
new file mode 100644
index 0000000000000000000000000000000000000000..f5c6285bfac3bf5e4ab92e4f1b4a780e928b74bb
--- /dev/null
+++ b/src/ptbench/data/tbx11k_simplified_v2/fold_3.py
@@ -0,0 +1,49 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""TBX11k simplified dataset for TB detection (cross validation fold 3)
+
+* Split reference: first 62.6% of CXR for "train", 16% for "validation",
+* 21.4% for "test"
+* This split consists of non-TB and active TB samples
+* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.tbx11k_simplified_v2` 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")
+ (
+ 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/tbx11k_simplified_v2/fold_3_rgb.py b/src/ptbench/data/tbx11k_simplified_v2/fold_3_rgb.py
new file mode 100644
index 0000000000000000000000000000000000000000..0aaab4af498453e58fec75a4dc887173c1494d46
--- /dev/null
+++ b/src/ptbench/data/tbx11k_simplified_v2/fold_3_rgb.py
@@ -0,0 +1,49 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""TBX11k simplified dataset for TB detection (cross validation fold 3, RGB)
+
+* Split reference: first 62.6% of CXR for "train", 16% for "validation",
+* 21.4% for "test"
+* This split consists of non-TB and active TB samples
+* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.tbx11k_simplified_v2` 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/tbx11k_simplified_v2/fold_4.py b/src/ptbench/data/tbx11k_simplified_v2/fold_4.py
new file mode 100644
index 0000000000000000000000000000000000000000..796cf8f6bc94562923d29d22a0d0299215682cb8
--- /dev/null
+++ b/src/ptbench/data/tbx11k_simplified_v2/fold_4.py
@@ -0,0 +1,49 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""TBX11k simplified dataset for TB detection (cross validation fold 4)
+
+* Split reference: first 62.6% of CXR for "train", 16% for "validation",
+* 21.4% for "test"
+* This split consists of non-TB and active TB samples
+* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.tbx11k_simplified_v2` 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")
+ (
+ 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/tbx11k_simplified_v2/fold_4_rgb.py b/src/ptbench/data/tbx11k_simplified_v2/fold_4_rgb.py
new file mode 100644
index 0000000000000000000000000000000000000000..b53968c6cc96d0735100a4928c1ce3d0e2ad51d3
--- /dev/null
+++ b/src/ptbench/data/tbx11k_simplified_v2/fold_4_rgb.py
@@ -0,0 +1,49 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""TBX11k simplified dataset for TB detection (cross validation fold 4, RGB)
+
+* Split reference: first 62.6% of CXR for "train", 16% for "validation",
+* 21.4% for "test"
+* This split consists of non-TB and active TB samples
+* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.tbx11k_simplified_v2` 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/tbx11k_simplified_v2/fold_5.py b/src/ptbench/data/tbx11k_simplified_v2/fold_5.py
new file mode 100644
index 0000000000000000000000000000000000000000..7adba117d4edf6984048c7321269d6d47163fa3b
--- /dev/null
+++ b/src/ptbench/data/tbx11k_simplified_v2/fold_5.py
@@ -0,0 +1,49 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""TBX11k simplified dataset for TB detection (cross validation fold 5)
+
+* Split reference: first 62.6% of CXR for "train", 16% for "validation",
+* 21.4% for "test"
+* This split consists of non-TB and active TB samples
+* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.tbx11k_simplified_v2` 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")
+ (
+ 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/tbx11k_simplified_v2/fold_5_rgb.py b/src/ptbench/data/tbx11k_simplified_v2/fold_5_rgb.py
new file mode 100644
index 0000000000000000000000000000000000000000..a7ae16472f80f32b83dc31e8ffb9ed746dd0ed78
--- /dev/null
+++ b/src/ptbench/data/tbx11k_simplified_v2/fold_5_rgb.py
@@ -0,0 +1,49 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""TBX11k simplified dataset for TB detection (cross validation fold 5, RGB)
+
+* Split reference: first 62.6% of CXR for "train", 16% for "validation",
+* 21.4% for "test"
+* This split consists of non-TB and active TB samples
+* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.tbx11k_simplified_v2` 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/tbx11k_simplified_v2/fold_6.py b/src/ptbench/data/tbx11k_simplified_v2/fold_6.py
new file mode 100644
index 0000000000000000000000000000000000000000..0a5199e1b9f21a2ddc058afe119ad816030e3a0b
--- /dev/null
+++ b/src/ptbench/data/tbx11k_simplified_v2/fold_6.py
@@ -0,0 +1,49 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""TBX11k simplified dataset for TB detection (cross validation fold 6)
+
+* Split reference: first 62.6% of CXR for "train", 16% for "validation",
+* 21.4% for "test"
+* This split consists of non-TB and active TB samples
+* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.tbx11k_simplified_v2` 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")
+ (
+ 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/tbx11k_simplified_v2/fold_6_rgb.py b/src/ptbench/data/tbx11k_simplified_v2/fold_6_rgb.py
new file mode 100644
index 0000000000000000000000000000000000000000..e46d97d57af3f676a6f390281f9d5e3cba8b8dbf
--- /dev/null
+++ b/src/ptbench/data/tbx11k_simplified_v2/fold_6_rgb.py
@@ -0,0 +1,49 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""TBX11k simplified dataset for TB detection (cross validation fold 6, RGB)
+
+* Split reference: first 62.6% of CXR for "train", 16% for "validation",
+* 21.4% for "test"
+* This split consists of non-TB and active TB samples
+* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.tbx11k_simplified_v2` 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/tbx11k_simplified_v2/fold_7.py b/src/ptbench/data/tbx11k_simplified_v2/fold_7.py
new file mode 100644
index 0000000000000000000000000000000000000000..f34d1d59883fd8ddf3c09eaac350b8782e45ca5f
--- /dev/null
+++ b/src/ptbench/data/tbx11k_simplified_v2/fold_7.py
@@ -0,0 +1,49 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""TBX11k simplified dataset for TB detection (cross validation fold 7)
+
+* Split reference: first 62.6% of CXR for "train", 16% for "validation",
+* 21.4% for "test"
+* This split consists of non-TB and active TB samples
+* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.tbx11k_simplified_v2` 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")
+ (
+ 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/tbx11k_simplified_v2/fold_7_rgb.py b/src/ptbench/data/tbx11k_simplified_v2/fold_7_rgb.py
new file mode 100644
index 0000000000000000000000000000000000000000..4c595e735fb47d63dacefc6b9eb4f3ef7fe9a9fc
--- /dev/null
+++ b/src/ptbench/data/tbx11k_simplified_v2/fold_7_rgb.py
@@ -0,0 +1,49 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""TBX11k simplified dataset for TB detection (cross validation fold 7, RGB)
+
+* Split reference: first 62.6% of CXR for "train", 16% for "validation",
+* 21.4% for "test"
+* This split consists of non-TB and active TB samples
+* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.tbx11k_simplified_v2` 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/tbx11k_simplified_v2/fold_8.py b/src/ptbench/data/tbx11k_simplified_v2/fold_8.py
new file mode 100644
index 0000000000000000000000000000000000000000..eb4a94aa48d8f7b4960ef134a5ad737ff0d88d4e
--- /dev/null
+++ b/src/ptbench/data/tbx11k_simplified_v2/fold_8.py
@@ -0,0 +1,49 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""TBX11k simplified dataset for TB detection (cross validation fold 8)
+
+* Split reference: first 62.6% of CXR for "train", 16% for "validation",
+* 21.4% for "test"
+* This split consists of non-TB and active TB samples
+* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.tbx11k_simplified_v2` 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")
+ (
+ 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/tbx11k_simplified_v2/fold_8_rgb.py b/src/ptbench/data/tbx11k_simplified_v2/fold_8_rgb.py
new file mode 100644
index 0000000000000000000000000000000000000000..70510ce5ee8c87cdff8190c0be875dd84e3d385c
--- /dev/null
+++ b/src/ptbench/data/tbx11k_simplified_v2/fold_8_rgb.py
@@ -0,0 +1,49 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""TBX11k simplified dataset for TB detection (cross validation fold 8, RGB)
+
+* Split reference: first 62.6% of CXR for "train", 16% for "validation",
+* 21.4% for "test"
+* This split consists of non-TB and active TB samples
+* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.tbx11k_simplified_v2` 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/tbx11k_simplified_v2/fold_9.py b/src/ptbench/data/tbx11k_simplified_v2/fold_9.py
new file mode 100644
index 0000000000000000000000000000000000000000..b15c1bf02990d9aae52a6c1c2b769ef54abf2321
--- /dev/null
+++ b/src/ptbench/data/tbx11k_simplified_v2/fold_9.py
@@ -0,0 +1,49 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""TBX11k simplified dataset for TB detection (cross validation fold 9)
+
+* Split reference: first 62.6% of CXR for "train", 16% for "validation",
+* 21.4% for "test"
+* This split consists of non-TB and active TB samples
+* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.tbx11k_simplified_v2` 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")
+ (
+ 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/tbx11k_simplified_v2/fold_9_rgb.py b/src/ptbench/data/tbx11k_simplified_v2/fold_9_rgb.py
new file mode 100644
index 0000000000000000000000000000000000000000..271e33480a84ff839cc997e3ae5b881e0595ae68
--- /dev/null
+++ b/src/ptbench/data/tbx11k_simplified_v2/fold_9_rgb.py
@@ -0,0 +1,49 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""TBX11k simplified dataset for TB detection (cross validation fold 9, RGB)
+
+* Split reference: first 62.6% of CXR for "train", 16% for "validation",
+* 21.4% for "test"
+* This split consists of non-TB and active TB samples
+* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.tbx11k_simplified_v2` 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/tbx11k_simplified_v2/rgb.py b/src/ptbench/data/tbx11k_simplified_v2/rgb.py
new file mode 100644
index 0000000000000000000000000000000000000000..13ac3a9f9cbdbbdb55c43026e54692909ac8299c
--- /dev/null
+++ b/src/ptbench/data/tbx11k_simplified_v2/rgb.py
@@ -0,0 +1,50 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""TBX11k simplified dataset for TB detection (default protocol, converted in
+RGB)
+
+* Split reference: first 62.6% of CXR for "train", 16% for "validation",
+* 21.4% for "test"
+* This split consists of non-TB and active TB samples
+* "healthy", "latent TB", and "sick & non-TB" samples are all merged under the label "non-TB"
+* This configuration resolution: 512 x 512 (default)
+* See :py:mod:`ptbench.data.tbx11k_simplified_v2` 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