diff --git a/pyproject.toml b/pyproject.toml
index 7c9f19488a8aa34e72d1faf2be868a255710654c..e6208923a93d82f69e13230ac9e4582474f0b504 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -365,28 +365,28 @@ mc_ch_in_11k_rs_f7 = "ptbench.configs.datasets.mc_ch_in_11k_RS.fold_7"
mc_ch_in_11k_rs_f8 = "ptbench.configs.datasets.mc_ch_in_11k_RS.fold_8"
mc_ch_in_11k_rs_f9 = "ptbench.configs.datasets.mc_ch_in_11k_RS.fold_9"
# montgomery-shenzhen-indian-tbx11kv2 aggregated dataset
-mc_ch_in_11kv2 = "ptbench.configs.datasets.mc_ch_in_11kv2.default"
-mc_ch_in_11kv2_rgb = "ptbench.configs.datasets.mc_ch_in_11kv2.rgb"
-mc_ch_in_11kv2_f0 = "ptbench.configs.datasets.mc_ch_in_11kv2.fold_0"
-mc_ch_in_11kv2_f1 = "ptbench.configs.datasets.mc_ch_in_11kv2.fold_1"
-mc_ch_in_11kv2_f2 = "ptbench.configs.datasets.mc_ch_in_11kv2.fold_2"
-mc_ch_in_11kv2_f3 = "ptbench.configs.datasets.mc_ch_in_11kv2.fold_3"
-mc_ch_in_11kv2_f4 = "ptbench.configs.datasets.mc_ch_in_11kv2.fold_4"
-mc_ch_in_11kv2_f5 = "ptbench.configs.datasets.mc_ch_in_11kv2.fold_5"
-mc_ch_in_11kv2_f6 = "ptbench.configs.datasets.mc_ch_in_11kv2.fold_6"
-mc_ch_in_11kv2_f7 = "ptbench.configs.datasets.mc_ch_in_11kv2.fold_7"
-mc_ch_in_11kv2_f8 = "ptbench.configs.datasets.mc_ch_in_11kv2.fold_8"
-mc_ch_in_11kv2_f9 = "ptbench.configs.datasets.mc_ch_in_11kv2.fold_9"
-mc_ch_in_11kv2_f0_rgb = "ptbench.configs.datasets.mc_ch_in_11kv2.fold_0_rgb"
-mc_ch_in_11kv2_f1_rgb = "ptbench.configs.datasets.mc_ch_in_11kv2.fold_1_rgb"
-mc_ch_in_11kv2_f2_rgb = "ptbench.configs.datasets.mc_ch_in_11kv2.fold_2_rgb"
-mc_ch_in_11kv2_f3_rgb = "ptbench.configs.datasets.mc_ch_in_11kv2.fold_3_rgb"
-mc_ch_in_11kv2_f4_rgb = "ptbench.configs.datasets.mc_ch_in_11kv2.fold_4_rgb"
-mc_ch_in_11kv2_f5_rgb = "ptbench.configs.datasets.mc_ch_in_11kv2.fold_5_rgb"
-mc_ch_in_11kv2_f6_rgb = "ptbench.configs.datasets.mc_ch_in_11kv2.fold_6_rgb"
-mc_ch_in_11kv2_f7_rgb = "ptbench.configs.datasets.mc_ch_in_11kv2.fold_7_rgb"
-mc_ch_in_11kv2_f8_rgb = "ptbench.configs.datasets.mc_ch_in_11kv2.fold_8_rgb"
-mc_ch_in_11kv2_f9_rgb = "ptbench.configs.datasets.mc_ch_in_11kv2.fold_9_rgb"
+mc_ch_in_11kv2 = "ptbench.data.mc_ch_in_11kv2.default"
+mc_ch_in_11kv2_rgb = "ptbench.data.mc_ch_in_11kv2.rgb"
+mc_ch_in_11kv2_f0 = "ptbench.data.mc_ch_in_11kv2.fold_0"
+mc_ch_in_11kv2_f1 = "ptbench.data.mc_ch_in_11kv2.fold_1"
+mc_ch_in_11kv2_f2 = "ptbench.data.mc_ch_in_11kv2.fold_2"
+mc_ch_in_11kv2_f3 = "ptbench.data.mc_ch_in_11kv2.fold_3"
+mc_ch_in_11kv2_f4 = "ptbench.data.mc_ch_in_11kv2.fold_4"
+mc_ch_in_11kv2_f5 = "ptbench.data.mc_ch_in_11kv2.fold_5"
+mc_ch_in_11kv2_f6 = "ptbench.data.mc_ch_in_11kv2.fold_6"
+mc_ch_in_11kv2_f7 = "ptbench.data.mc_ch_in_11kv2.fold_7"
+mc_ch_in_11kv2_f8 = "ptbench.data.mc_ch_in_11kv2.fold_8"
+mc_ch_in_11kv2_f9 = "ptbench.data.mc_ch_in_11kv2.fold_9"
+mc_ch_in_11kv2_f0_rgb = "ptbench.data.mc_ch_in_11kv2.fold_0_rgb"
+mc_ch_in_11kv2_f1_rgb = "ptbench.data.mc_ch_in_11kv2.fold_1_rgb"
+mc_ch_in_11kv2_f2_rgb = "ptbench.data.mc_ch_in_11kv2.fold_2_rgb"
+mc_ch_in_11kv2_f3_rgb = "ptbench.data.mc_ch_in_11kv2.fold_3_rgb"
+mc_ch_in_11kv2_f4_rgb = "ptbench.data.mc_ch_in_11kv2.fold_4_rgb"
+mc_ch_in_11kv2_f5_rgb = "ptbench.data.mc_ch_in_11kv2.fold_5_rgb"
+mc_ch_in_11kv2_f6_rgb = "ptbench.data.mc_ch_in_11kv2.fold_6_rgb"
+mc_ch_in_11kv2_f7_rgb = "ptbench.data.mc_ch_in_11kv2.fold_7_rgb"
+mc_ch_in_11kv2_f8_rgb = "ptbench.data.mc_ch_in_11kv2.fold_8_rgb"
+mc_ch_in_11kv2_f9_rgb = "ptbench.data.mc_ch_in_11kv2.fold_9_rgb"
# extended montgomery-shenzhen-indian-tbx11kv2 aggregated dataset
# (with radiological signs)
mc_ch_in_11kv2_rs = "ptbench.configs.datasets.mc_ch_in_11kv2_RS.default"
diff --git a/src/ptbench/configs/datasets/mc_ch_in_11kv2/__init__.py b/src/ptbench/configs/datasets/mc_ch_in_11kv2/__init__.py
deleted file mode 100644
index c36f7f60c8d1111baa7f3559c419e134b7ece62f..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/mc_ch_in_11kv2/__init__.py
+++ /dev/null
@@ -1,157 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-from torch.utils.data.dataset import ConcatDataset
-
-
-def _maker(protocol):
- if protocol == "default":
- from ..indian import default as indian
- from ..montgomery import default as mc
- from ..shenzhen import default as ch
- from ..tbx11k_simplified_v2 import default as tbx11kv2
- elif protocol == "rgb":
- from ..indian import rgb as indian
- from ..montgomery import rgb as mc
- from ..shenzhen import rgb as ch
- from ..tbx11k_simplified_v2 import rgb as tbx11kv2
- elif protocol == "fold_0":
- from ..indian import fold_0 as indian
- from ..montgomery import fold_0 as mc
- from ..shenzhen import fold_0 as ch
- from ..tbx11k_simplified_v2 import fold_0 as tbx11kv2
- elif protocol == "fold_1":
- from ..indian import fold_1 as indian
- from ..montgomery import fold_1 as mc
- from ..shenzhen import fold_1 as ch
- from ..tbx11k_simplified_v2 import fold_1 as tbx11kv2
- elif protocol == "fold_2":
- from ..indian import fold_2 as indian
- from ..montgomery import fold_2 as mc
- from ..shenzhen import fold_2 as ch
- from ..tbx11k_simplified_v2 import fold_2 as tbx11kv2
- elif protocol == "fold_3":
- from ..indian import fold_3 as indian
- from ..montgomery import fold_3 as mc
- from ..shenzhen import fold_3 as ch
- from ..tbx11k_simplified_v2 import fold_3 as tbx11kv2
- elif protocol == "fold_4":
- from ..indian import fold_4 as indian
- from ..montgomery import fold_4 as mc
- from ..shenzhen import fold_4 as ch
- from ..tbx11k_simplified_v2 import fold_4 as tbx11kv2
- elif protocol == "fold_5":
- from ..indian import fold_5 as indian
- from ..montgomery import fold_5 as mc
- from ..shenzhen import fold_5 as ch
- from ..tbx11k_simplified_v2 import fold_5 as tbx11kv2
- elif protocol == "fold_6":
- from ..indian import fold_6 as indian
- from ..montgomery import fold_6 as mc
- from ..shenzhen import fold_6 as ch
- from ..tbx11k_simplified_v2 import fold_6 as tbx11kv2
- elif protocol == "fold_7":
- from ..indian import fold_7 as indian
- from ..montgomery import fold_7 as mc
- from ..shenzhen import fold_7 as ch
- from ..tbx11k_simplified_v2 import fold_7 as tbx11kv2
- elif protocol == "fold_8":
- from ..indian import fold_8 as indian
- from ..montgomery import fold_8 as mc
- from ..shenzhen import fold_8 as ch
- from ..tbx11k_simplified_v2 import fold_8 as tbx11kv2
- elif protocol == "fold_9":
- from ..indian import fold_9 as indian
- from ..montgomery import fold_9 as mc
- from ..shenzhen import fold_9 as ch
- from ..tbx11k_simplified_v2 import fold_9 as tbx11kv2
- elif protocol == "fold_0_rgb":
- from ..indian import fold_0_rgb as indian
- from ..montgomery import fold_0_rgb as mc
- from ..shenzhen import fold_0_rgb as ch
- from ..tbx11k_simplified_v2 import fold_0_rgb as tbx11kv2
- elif protocol == "fold_1_rgb":
- from ..indian import fold_1_rgb as indian
- from ..montgomery import fold_1_rgb as mc
- from ..shenzhen import fold_1_rgb as ch
- from ..tbx11k_simplified_v2 import fold_1_rgb as tbx11kv2
- elif protocol == "fold_2_rgb":
- from ..indian import fold_2_rgb as indian
- from ..montgomery import fold_2_rgb as mc
- from ..shenzhen import fold_2_rgb as ch
- from ..tbx11k_simplified_v2 import fold_2_rgb as tbx11kv2
- elif protocol == "fold_3_rgb":
- from ..indian import fold_3_rgb as indian
- from ..montgomery import fold_3_rgb as mc
- from ..shenzhen import fold_3_rgb as ch
- from ..tbx11k_simplified_v2 import fold_3_rgb as tbx11kv2
- elif protocol == "fold_4_rgb":
- from ..indian import fold_4_rgb as indian
- from ..montgomery import fold_4_rgb as mc
- from ..shenzhen import fold_4_rgb as ch
- from ..tbx11k_simplified_v2 import fold_4_rgb as tbx11kv2
- elif protocol == "fold_5_rgb":
- from ..indian import fold_5_rgb as indian
- from ..montgomery import fold_5_rgb as mc
- from ..shenzhen import fold_5_rgb as ch
- from ..tbx11k_simplified_v2 import fold_5_rgb as tbx11kv2
- elif protocol == "fold_6_rgb":
- from ..indian import fold_6_rgb as indian
- from ..montgomery import fold_6_rgb as mc
- from ..shenzhen import fold_6_rgb as ch
- from ..tbx11k_simplified_v2 import fold_6_rgb as tbx11kv2
- elif protocol == "fold_7_rgb":
- from ..indian import fold_7_rgb as indian
- from ..montgomery import fold_7_rgb as mc
- from ..shenzhen import fold_7_rgb as ch
- from ..tbx11k_simplified_v2 import fold_7_rgb as tbx11kv2
- elif protocol == "fold_8_rgb":
- from ..indian import fold_8_rgb as indian
- from ..montgomery import fold_8_rgb as mc
- from ..shenzhen import fold_8_rgb as ch
- from ..tbx11k_simplified_v2 import fold_8_rgb as tbx11kv2
- elif protocol == "fold_9_rgb":
- from ..indian import fold_9_rgb as indian
- from ..montgomery import fold_9_rgb as mc
- from ..shenzhen import fold_9_rgb as ch
- from ..tbx11k_simplified_v2 import fold_9_rgb as tbx11kv2
-
- mc = mc.dataset
- ch = ch.dataset
- indian = indian.dataset
- tbx11kv2 = tbx11kv2.dataset
-
- dataset = {}
- dataset["__train__"] = ConcatDataset(
- [
- mc["__train__"],
- ch["__train__"],
- indian["__train__"],
- tbx11kv2["__train__"],
- ]
- )
- dataset["train"] = ConcatDataset(
- [mc["train"], ch["train"], indian["train"], tbx11kv2["train"]]
- )
- dataset["__valid__"] = ConcatDataset(
- [
- mc["__valid__"],
- ch["__valid__"],
- indian["__valid__"],
- tbx11kv2["__valid__"],
- ]
- )
- dataset["validation"] = ConcatDataset(
- [
- mc["validation"],
- ch["validation"],
- indian["validation"],
- tbx11kv2["validation"],
- ]
- )
- dataset["test"] = ConcatDataset(
- [mc["test"], ch["test"], indian["test"], tbx11kv2["test"]]
- )
-
- return dataset
diff --git a/src/ptbench/configs/datasets/mc_ch_in_11kv2/default.py b/src/ptbench/configs/datasets/mc_ch_in_11kv2/default.py
deleted file mode 100644
index 7d4f16bda48b05e7e9302ffc9c689d8393b3e495..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/mc_ch_in_11kv2/default.py
+++ /dev/null
@@ -1,10 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
-TBX11K-simplified datasets."""
-
-from . import _maker
-
-dataset = _maker("default")
diff --git a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_0.py b/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_0.py
deleted file mode 100644
index 757a0eb98214ba020d76095363d424b9209540e7..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_0.py
+++ /dev/null
@@ -1,10 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
-TBX11K-simplified datasets (cross validation fold 0)"""
-
-from . import _maker
-
-dataset = _maker("fold_0")
diff --git a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_0_rgb.py b/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_0_rgb.py
deleted file mode 100644
index 48e05ff3f71f13976190d04cfaf59c5c36996bac..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_0_rgb.py
+++ /dev/null
@@ -1,10 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
-TBX11K-simplified datasets (cross validation fold 0, RGB)"""
-
-from . import _maker
-
-dataset = _maker("fold_0_rgb")
diff --git a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_1.py b/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_1.py
deleted file mode 100644
index 5657958934b926879bd26503c9b383e775bc724d..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_1.py
+++ /dev/null
@@ -1,10 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
-TBX11K-simplified datasets (cross validation fold 1)"""
-
-from . import _maker
-
-dataset = _maker("fold_1")
diff --git a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_1_rgb.py b/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_1_rgb.py
deleted file mode 100644
index c782d68de247c876ddd6826100cbb7908342b928..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_1_rgb.py
+++ /dev/null
@@ -1,10 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
-TBX11K-simplified datasets (cross validation fold 1, RGB)"""
-
-from . import _maker
-
-dataset = _maker("fold_1_rgb")
diff --git a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_2.py b/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_2.py
deleted file mode 100644
index 10a597bcb8e0485db63f0d7500b15b3e78877066..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_2.py
+++ /dev/null
@@ -1,10 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
-TBX11K-simplified datasets (cross validation fold 2)"""
-
-from . import _maker
-
-dataset = _maker("fold_2")
diff --git a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_2_rgb.py b/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_2_rgb.py
deleted file mode 100644
index d624f3af53abcf053c7bf17a9822a86cb53e2923..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_2_rgb.py
+++ /dev/null
@@ -1,10 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
-TBX11K-simplified datasets (cross validation fold 2, RGB)"""
-
-from . import _maker
-
-dataset = _maker("fold_2_rgb")
diff --git a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_3.py b/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_3.py
deleted file mode 100644
index 39bee4fec99e81eecc22a365183283bcd2ec3d98..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_3.py
+++ /dev/null
@@ -1,10 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
-TBX11K-simplified datasets (cross validation fold 3)"""
-
-from . import _maker
-
-dataset = _maker("fold_3")
diff --git a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_3_rgb.py b/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_3_rgb.py
deleted file mode 100644
index 7b26e4257e61013843e3a62c3bc419003e23b645..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_3_rgb.py
+++ /dev/null
@@ -1,10 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
-TBX11K-simplified datasets (cross validation fold 3, RGB)"""
-
-from . import _maker
-
-dataset = _maker("fold_3_rgb")
diff --git a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_4.py b/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_4.py
deleted file mode 100644
index 5fb56292fd97636f452cde06c87bb34c89f01b1c..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_4.py
+++ /dev/null
@@ -1,10 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
-TBX11K-simplified datasets (cross validation fold 4)"""
-
-from . import _maker
-
-dataset = _maker("fold_4")
diff --git a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_4_rgb.py b/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_4_rgb.py
deleted file mode 100644
index fbc4f0cfd9edc602fbe5665aca0465b29c5183b5..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_4_rgb.py
+++ /dev/null
@@ -1,10 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
-TBX11K-simplified datasets (cross validation fold 4, RGB)"""
-
-from . import _maker
-
-dataset = _maker("fold_4_rgb")
diff --git a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_5.py b/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_5.py
deleted file mode 100644
index 679bb9b3cbbdede06cd87834239609720f439296..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_5.py
+++ /dev/null
@@ -1,10 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
-TBX11K-simplified datasets (cross validation fold 5)"""
-
-from . import _maker
-
-dataset = _maker("fold_5")
diff --git a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_5_rgb.py b/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_5_rgb.py
deleted file mode 100644
index 747d510ecd1c7bd2f32ab7b139a53603d5bbee88..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_5_rgb.py
+++ /dev/null
@@ -1,10 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
-TBX11K-simplified datasets (cross validation fold 5, RGB)"""
-
-from . import _maker
-
-dataset = _maker("fold_5_rgb")
diff --git a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_6.py b/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_6.py
deleted file mode 100644
index cd8e4cd571b8c796bad3221584870888c5186d3d..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_6.py
+++ /dev/null
@@ -1,10 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
-TBX11K-simplified datasets (cross validation fold 6)"""
-
-from . import _maker
-
-dataset = _maker("fold_6")
diff --git a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_6_rgb.py b/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_6_rgb.py
deleted file mode 100644
index 86f112c3aae0c1c1dd48002347f78ce565797d47..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_6_rgb.py
+++ /dev/null
@@ -1,10 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
-TBX11K-simplified datasets (cross validation fold 6, RGB)"""
-
-from . import _maker
-
-dataset = _maker("fold_6_rgb")
diff --git a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_7.py b/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_7.py
deleted file mode 100644
index 98241531d3e15720f07ef9174687c47db7d737f1..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_7.py
+++ /dev/null
@@ -1,10 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
-TBX11K-simplified datasets (cross validation fold 7)"""
-
-from . import _maker
-
-dataset = _maker("fold_7")
diff --git a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_7_rgb.py b/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_7_rgb.py
deleted file mode 100644
index 981fe19180e0d8d4e1b21653f52a92a567723a63..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_7_rgb.py
+++ /dev/null
@@ -1,10 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
-TBX11K-simplified datasets (cross validation fold 7, RGB)"""
-
-from . import _maker
-
-dataset = _maker("fold_7_rgb")
diff --git a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_8.py b/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_8.py
deleted file mode 100644
index dab1a234a3842ab450706d86060651d4383ddbfc..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_8.py
+++ /dev/null
@@ -1,10 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
-TBX11K-simplified datasets (cross validation fold 8)"""
-
-from . import _maker
-
-dataset = _maker("fold_8")
diff --git a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_8_rgb.py b/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_8_rgb.py
deleted file mode 100644
index 798b8de64761ef0d87f491ef08b43426f55898f2..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_8_rgb.py
+++ /dev/null
@@ -1,10 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
-TBX11K-simplified datasets (cross validation fold 8, RGB)"""
-
-from . import _maker
-
-dataset = _maker("fold_8_rgb")
diff --git a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_9.py b/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_9.py
deleted file mode 100644
index 097724b9446c4c2f0bef8ee6f838c1c11ff627a5..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_9.py
+++ /dev/null
@@ -1,10 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
-TBX11K-simplified datasets (cross validation fold 9)"""
-
-from . import _maker
-
-dataset = _maker("fold_9")
diff --git a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_9_rgb.py b/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_9_rgb.py
deleted file mode 100644
index c6c564a40b957b562a37bb30b5809f7cf680e896..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/mc_ch_in_11kv2/fold_9_rgb.py
+++ /dev/null
@@ -1,10 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
-TBX11K-simplified datasets (cross validation fold 9, RGB)"""
-
-from . import _maker
-
-dataset = _maker("fold_9_rgb")
diff --git a/src/ptbench/configs/datasets/mc_ch_in_11kv2/rgb.py b/src/ptbench/configs/datasets/mc_ch_in_11kv2/rgb.py
deleted file mode 100644
index f47796a89c31a5a31c0f972d81b5d97c7f8742b4..0000000000000000000000000000000000000000
--- a/src/ptbench/configs/datasets/mc_ch_in_11kv2/rgb.py
+++ /dev/null
@@ -1,10 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-
-"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
-TBX11K-simplified datasets (RGB)"""
-
-from . import _maker
-
-dataset = _maker("rgb")
diff --git a/src/ptbench/data/mc_ch_in_11kv2/__init__.py b/src/ptbench/data/mc_ch_in_11kv2/__init__.py
new file mode 100644
index 0000000000000000000000000000000000000000..662d5c1326651b4d9f48d47bc4b503df23d17216
--- /dev/null
+++ b/src/ptbench/data/mc_ch_in_11kv2/__init__.py
@@ -0,0 +1,3 @@
+# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
diff --git a/src/ptbench/data/mc_ch_in_11kv2/default.py b/src/ptbench/data/mc_ch_in_11kv2/default.py
new file mode 100644
index 0000000000000000000000000000000000000000..1eea460b5503e79c755d52d2e5f08fd28d7ed3f7
--- /dev/null
+++ b/src/ptbench/data/mc_ch_in_11kv2/default.py
@@ -0,0 +1,101 @@
+# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
+TBX11K-simplified datasets."""
+
+from clapper.logging import setup
+from torch.utils.data.dataset import ConcatDataset
+
+from .. import return_subsets
+from ..base_datamodule import BaseDataModule, get_dataset_from_module
+from ..indian.default import datamodule as indian_datamodule
+from ..montgomery.default import datamodule as mc_datamodule
+from ..shenzhen.default import datamodule as ch_datamodule
+from ..tbx11k_simplified_v2.default import datamodule as tbx11kv2_datamodule
+
+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,
+ ):
+ self.train_batch_size = train_batch_size
+ self.predict_batch_size = predict_batch_size
+ self.drop_incomplete_batch = drop_incomplete_batch
+ self.multiproc_kwargs = multiproc_kwargs
+
+ 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):
+ # Instantiate other datamodules and get their datasets
+
+ module_args = {
+ "train_batch_size": self.train_batch_size,
+ "predict_batch_size": self.predict_batch_size,
+ "drop_incomplete_batch": self.drop_incomplete_batch,
+ "multiproc_kwargs": self.multiproc_kwargs,
+ }
+
+ mc = get_dataset_from_module(mc_datamodule, stage, **module_args)
+ ch = get_dataset_from_module(ch_datamodule, stage, **module_args)
+ indian = get_dataset_from_module(
+ indian_datamodule, stage, **module_args
+ )
+ tbx11kv2 = get_dataset_from_module(
+ tbx11kv2_datamodule, stage, **module_args
+ )
+
+ # Combine datasets
+ self.dataset = {}
+ self.dataset["__train__"] = ConcatDataset(
+ [
+ mc["__train__"],
+ ch["__train__"],
+ indian["__train__"],
+ tbx11kv2["__train__"],
+ ]
+ )
+ self.dataset["train"] = ConcatDataset(
+ [mc["train"], ch["train"], indian["train"], tbx11kv2["train"]]
+ )
+ self.dataset["__valid__"] = ConcatDataset(
+ [
+ mc["__valid__"],
+ ch["__valid__"],
+ indian["__valid__"],
+ tbx11kv2["__valid__"],
+ ]
+ )
+ self.dataset["validation"] = ConcatDataset(
+ [
+ mc["validation"],
+ ch["validation"],
+ indian["validation"],
+ tbx11kv2["validation"],
+ ]
+ )
+ self.dataset["test"] = ConcatDataset(
+ [mc["test"], ch["test"], indian["test"], tbx11kv2["test"]]
+ )
+
+ (
+ 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/mc_ch_in_11kv2/fold_0.py b/src/ptbench/data/mc_ch_in_11kv2/fold_0.py
new file mode 100644
index 0000000000000000000000000000000000000000..738164a78835fa7bed8add6a9e5e0fe101abad52
--- /dev/null
+++ b/src/ptbench/data/mc_ch_in_11kv2/fold_0.py
@@ -0,0 +1,101 @@
+# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
+TBX11K-simplified datasets (cross validation fold 0)"""
+
+from clapper.logging import setup
+from torch.utils.data.dataset import ConcatDataset
+
+from .. import return_subsets
+from ..base_datamodule import BaseDataModule, get_dataset_from_module
+from ..indian.fold_0 import datamodule as indian_datamodule
+from ..montgomery.fold_0 import datamodule as mc_datamodule
+from ..shenzhen.fold_0 import datamodule as ch_datamodule
+from ..tbx11k_simplified_v2.fold_0 import datamodule as tbx11kv2_datamodule
+
+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,
+ ):
+ self.train_batch_size = train_batch_size
+ self.predict_batch_size = predict_batch_size
+ self.drop_incomplete_batch = drop_incomplete_batch
+ self.multiproc_kwargs = multiproc_kwargs
+
+ 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):
+ # Instantiate other datamodules and get their datasets
+
+ module_args = {
+ "train_batch_size": self.train_batch_size,
+ "predict_batch_size": self.predict_batch_size,
+ "drop_incomplete_batch": self.drop_incomplete_batch,
+ "multiproc_kwargs": self.multiproc_kwargs,
+ }
+
+ mc = get_dataset_from_module(mc_datamodule, stage, **module_args)
+ ch = get_dataset_from_module(ch_datamodule, stage, **module_args)
+ indian = get_dataset_from_module(
+ indian_datamodule, stage, **module_args
+ )
+ tbx11kv2 = get_dataset_from_module(
+ tbx11kv2_datamodule, stage, **module_args
+ )
+
+ # Combine datasets
+ self.dataset = {}
+ self.dataset["__train__"] = ConcatDataset(
+ [
+ mc["__train__"],
+ ch["__train__"],
+ indian["__train__"],
+ tbx11kv2["__train__"],
+ ]
+ )
+ self.dataset["train"] = ConcatDataset(
+ [mc["train"], ch["train"], indian["train"], tbx11kv2["train"]]
+ )
+ self.dataset["__valid__"] = ConcatDataset(
+ [
+ mc["__valid__"],
+ ch["__valid__"],
+ indian["__valid__"],
+ tbx11kv2["__valid__"],
+ ]
+ )
+ self.dataset["validation"] = ConcatDataset(
+ [
+ mc["validation"],
+ ch["validation"],
+ indian["validation"],
+ tbx11kv2["validation"],
+ ]
+ )
+ self.dataset["test"] = ConcatDataset(
+ [mc["test"], ch["test"], indian["test"], tbx11kv2["test"]]
+ )
+
+ (
+ 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/mc_ch_in_11kv2/fold_0_rgb.py b/src/ptbench/data/mc_ch_in_11kv2/fold_0_rgb.py
new file mode 100644
index 0000000000000000000000000000000000000000..c852a9a92ee5faf6a2accf83c3faa4169e829447
--- /dev/null
+++ b/src/ptbench/data/mc_ch_in_11kv2/fold_0_rgb.py
@@ -0,0 +1,101 @@
+# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
+TBX11K-simplified datasets (cross validation fold 0, RGB)"""
+
+from clapper.logging import setup
+from torch.utils.data.dataset import ConcatDataset
+
+from .. import return_subsets
+from ..base_datamodule import BaseDataModule, get_dataset_from_module
+from ..indian.fold_0_rgb import datamodule as indian_datamodule
+from ..montgomery.fold_0_rgb import datamodule as mc_datamodule
+from ..shenzhen.fold_0_rgb import datamodule as ch_datamodule
+from ..tbx11k_simplified_v2.fold_0_rgb import datamodule as tbx11kv2_datamodule
+
+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,
+ ):
+ self.train_batch_size = train_batch_size
+ self.predict_batch_size = predict_batch_size
+ self.drop_incomplete_batch = drop_incomplete_batch
+ self.multiproc_kwargs = multiproc_kwargs
+
+ 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):
+ # Instantiate other datamodules and get their datasets
+
+ module_args = {
+ "train_batch_size": self.train_batch_size,
+ "predict_batch_size": self.predict_batch_size,
+ "drop_incomplete_batch": self.drop_incomplete_batch,
+ "multiproc_kwargs": self.multiproc_kwargs,
+ }
+
+ mc = get_dataset_from_module(mc_datamodule, stage, **module_args)
+ ch = get_dataset_from_module(ch_datamodule, stage, **module_args)
+ indian = get_dataset_from_module(
+ indian_datamodule, stage, **module_args
+ )
+ tbx11kv2 = get_dataset_from_module(
+ tbx11kv2_datamodule, stage, **module_args
+ )
+
+ # Combine datasets
+ self.dataset = {}
+ self.dataset["__train__"] = ConcatDataset(
+ [
+ mc["__train__"],
+ ch["__train__"],
+ indian["__train__"],
+ tbx11kv2["__train__"],
+ ]
+ )
+ self.dataset["train"] = ConcatDataset(
+ [mc["train"], ch["train"], indian["train"], tbx11kv2["train"]]
+ )
+ self.dataset["__valid__"] = ConcatDataset(
+ [
+ mc["__valid__"],
+ ch["__valid__"],
+ indian["__valid__"],
+ tbx11kv2["__valid__"],
+ ]
+ )
+ self.dataset["validation"] = ConcatDataset(
+ [
+ mc["validation"],
+ ch["validation"],
+ indian["validation"],
+ tbx11kv2["validation"],
+ ]
+ )
+ self.dataset["test"] = ConcatDataset(
+ [mc["test"], ch["test"], indian["test"], tbx11kv2["test"]]
+ )
+
+ (
+ 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/mc_ch_in_11kv2/fold_1.py b/src/ptbench/data/mc_ch_in_11kv2/fold_1.py
new file mode 100644
index 0000000000000000000000000000000000000000..bfc2dbfc53a33325549f4cb41513201a005bb1f9
--- /dev/null
+++ b/src/ptbench/data/mc_ch_in_11kv2/fold_1.py
@@ -0,0 +1,101 @@
+# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
+TBX11K-simplified datasets (cross validation fold 1)"""
+
+from clapper.logging import setup
+from torch.utils.data.dataset import ConcatDataset
+
+from .. import return_subsets
+from ..base_datamodule import BaseDataModule, get_dataset_from_module
+from ..indian.fold_1 import datamodule as indian_datamodule
+from ..montgomery.fold_1 import datamodule as mc_datamodule
+from ..shenzhen.fold_1 import datamodule as ch_datamodule
+from ..tbx11k_simplified_v2.fold_1 import datamodule as tbx11kv2_datamodule
+
+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,
+ ):
+ self.train_batch_size = train_batch_size
+ self.predict_batch_size = predict_batch_size
+ self.drop_incomplete_batch = drop_incomplete_batch
+ self.multiproc_kwargs = multiproc_kwargs
+
+ 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):
+ # Instantiate other datamodules and get their datasets
+
+ module_args = {
+ "train_batch_size": self.train_batch_size,
+ "predict_batch_size": self.predict_batch_size,
+ "drop_incomplete_batch": self.drop_incomplete_batch,
+ "multiproc_kwargs": self.multiproc_kwargs,
+ }
+
+ mc = get_dataset_from_module(mc_datamodule, stage, **module_args)
+ ch = get_dataset_from_module(ch_datamodule, stage, **module_args)
+ indian = get_dataset_from_module(
+ indian_datamodule, stage, **module_args
+ )
+ tbx11kv2 = get_dataset_from_module(
+ tbx11kv2_datamodule, stage, **module_args
+ )
+
+ # Combine datasets
+ self.dataset = {}
+ self.dataset["__train__"] = ConcatDataset(
+ [
+ mc["__train__"],
+ ch["__train__"],
+ indian["__train__"],
+ tbx11kv2["__train__"],
+ ]
+ )
+ self.dataset["train"] = ConcatDataset(
+ [mc["train"], ch["train"], indian["train"], tbx11kv2["train"]]
+ )
+ self.dataset["__valid__"] = ConcatDataset(
+ [
+ mc["__valid__"],
+ ch["__valid__"],
+ indian["__valid__"],
+ tbx11kv2["__valid__"],
+ ]
+ )
+ self.dataset["validation"] = ConcatDataset(
+ [
+ mc["validation"],
+ ch["validation"],
+ indian["validation"],
+ tbx11kv2["validation"],
+ ]
+ )
+ self.dataset["test"] = ConcatDataset(
+ [mc["test"], ch["test"], indian["test"], tbx11kv2["test"]]
+ )
+
+ (
+ 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/mc_ch_in_11kv2/fold_1_rgb.py b/src/ptbench/data/mc_ch_in_11kv2/fold_1_rgb.py
new file mode 100644
index 0000000000000000000000000000000000000000..8f0c9174ffb8a7307fed77996fe0bbad655ad34c
--- /dev/null
+++ b/src/ptbench/data/mc_ch_in_11kv2/fold_1_rgb.py
@@ -0,0 +1,101 @@
+# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
+TBX11K-simplified datasets (cross validation fold 1, RGB)"""
+
+from clapper.logging import setup
+from torch.utils.data.dataset import ConcatDataset
+
+from .. import return_subsets
+from ..base_datamodule import BaseDataModule, get_dataset_from_module
+from ..indian.fold_1_rgb import datamodule as indian_datamodule
+from ..montgomery.fold_1_rgb import datamodule as mc_datamodule
+from ..shenzhen.fold_1_rgb import datamodule as ch_datamodule
+from ..tbx11k_simplified_v2.fold_1_rgb import datamodule as tbx11kv2_datamodule
+
+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,
+ ):
+ self.train_batch_size = train_batch_size
+ self.predict_batch_size = predict_batch_size
+ self.drop_incomplete_batch = drop_incomplete_batch
+ self.multiproc_kwargs = multiproc_kwargs
+
+ 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):
+ # Instantiate other datamodules and get their datasets
+
+ module_args = {
+ "train_batch_size": self.train_batch_size,
+ "predict_batch_size": self.predict_batch_size,
+ "drop_incomplete_batch": self.drop_incomplete_batch,
+ "multiproc_kwargs": self.multiproc_kwargs,
+ }
+
+ mc = get_dataset_from_module(mc_datamodule, stage, **module_args)
+ ch = get_dataset_from_module(ch_datamodule, stage, **module_args)
+ indian = get_dataset_from_module(
+ indian_datamodule, stage, **module_args
+ )
+ tbx11kv2 = get_dataset_from_module(
+ tbx11kv2_datamodule, stage, **module_args
+ )
+
+ # Combine datasets
+ self.dataset = {}
+ self.dataset["__train__"] = ConcatDataset(
+ [
+ mc["__train__"],
+ ch["__train__"],
+ indian["__train__"],
+ tbx11kv2["__train__"],
+ ]
+ )
+ self.dataset["train"] = ConcatDataset(
+ [mc["train"], ch["train"], indian["train"], tbx11kv2["train"]]
+ )
+ self.dataset["__valid__"] = ConcatDataset(
+ [
+ mc["__valid__"],
+ ch["__valid__"],
+ indian["__valid__"],
+ tbx11kv2["__valid__"],
+ ]
+ )
+ self.dataset["validation"] = ConcatDataset(
+ [
+ mc["validation"],
+ ch["validation"],
+ indian["validation"],
+ tbx11kv2["validation"],
+ ]
+ )
+ self.dataset["test"] = ConcatDataset(
+ [mc["test"], ch["test"], indian["test"], tbx11kv2["test"]]
+ )
+
+ (
+ 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/mc_ch_in_11kv2/fold_2.py b/src/ptbench/data/mc_ch_in_11kv2/fold_2.py
new file mode 100644
index 0000000000000000000000000000000000000000..06c7c1ab9b9ca28a6cdbf5e95108a669349a32a6
--- /dev/null
+++ b/src/ptbench/data/mc_ch_in_11kv2/fold_2.py
@@ -0,0 +1,101 @@
+# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
+TBX11K-simplified datasets (cross validation fold 2)"""
+
+from clapper.logging import setup
+from torch.utils.data.dataset import ConcatDataset
+
+from .. import return_subsets
+from ..base_datamodule import BaseDataModule, get_dataset_from_module
+from ..indian.fold_2 import datamodule as indian_datamodule
+from ..montgomery.fold_2 import datamodule as mc_datamodule
+from ..shenzhen.fold_2 import datamodule as ch_datamodule
+from ..tbx11k_simplified_v2.fold_2 import datamodule as tbx11kv2_datamodule
+
+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,
+ ):
+ self.train_batch_size = train_batch_size
+ self.predict_batch_size = predict_batch_size
+ self.drop_incomplete_batch = drop_incomplete_batch
+ self.multiproc_kwargs = multiproc_kwargs
+
+ 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):
+ # Instantiate other datamodules and get their datasets
+
+ module_args = {
+ "train_batch_size": self.train_batch_size,
+ "predict_batch_size": self.predict_batch_size,
+ "drop_incomplete_batch": self.drop_incomplete_batch,
+ "multiproc_kwargs": self.multiproc_kwargs,
+ }
+
+ mc = get_dataset_from_module(mc_datamodule, stage, **module_args)
+ ch = get_dataset_from_module(ch_datamodule, stage, **module_args)
+ indian = get_dataset_from_module(
+ indian_datamodule, stage, **module_args
+ )
+ tbx11kv2 = get_dataset_from_module(
+ tbx11kv2_datamodule, stage, **module_args
+ )
+
+ # Combine datasets
+ self.dataset = {}
+ self.dataset["__train__"] = ConcatDataset(
+ [
+ mc["__train__"],
+ ch["__train__"],
+ indian["__train__"],
+ tbx11kv2["__train__"],
+ ]
+ )
+ self.dataset["train"] = ConcatDataset(
+ [mc["train"], ch["train"], indian["train"], tbx11kv2["train"]]
+ )
+ self.dataset["__valid__"] = ConcatDataset(
+ [
+ mc["__valid__"],
+ ch["__valid__"],
+ indian["__valid__"],
+ tbx11kv2["__valid__"],
+ ]
+ )
+ self.dataset["validation"] = ConcatDataset(
+ [
+ mc["validation"],
+ ch["validation"],
+ indian["validation"],
+ tbx11kv2["validation"],
+ ]
+ )
+ self.dataset["test"] = ConcatDataset(
+ [mc["test"], ch["test"], indian["test"], tbx11kv2["test"]]
+ )
+
+ (
+ 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/mc_ch_in_11kv2/fold_2_rgb.py b/src/ptbench/data/mc_ch_in_11kv2/fold_2_rgb.py
new file mode 100644
index 0000000000000000000000000000000000000000..5ecaa2a7e5a8ae7972a06eb1fa68c39cb575bba3
--- /dev/null
+++ b/src/ptbench/data/mc_ch_in_11kv2/fold_2_rgb.py
@@ -0,0 +1,101 @@
+# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
+TBX11K-simplified datasets (cross validation fold 2, RGB)"""
+
+from clapper.logging import setup
+from torch.utils.data.dataset import ConcatDataset
+
+from .. import return_subsets
+from ..base_datamodule import BaseDataModule, get_dataset_from_module
+from ..indian.fold_2_rgb import datamodule as indian_datamodule
+from ..montgomery.fold_2_rgb import datamodule as mc_datamodule
+from ..shenzhen.fold_2_rgb import datamodule as ch_datamodule
+from ..tbx11k_simplified_v2.fold_2_rgb import datamodule as tbx11kv2_datamodule
+
+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,
+ ):
+ self.train_batch_size = train_batch_size
+ self.predict_batch_size = predict_batch_size
+ self.drop_incomplete_batch = drop_incomplete_batch
+ self.multiproc_kwargs = multiproc_kwargs
+
+ 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):
+ # Instantiate other datamodules and get their datasets
+
+ module_args = {
+ "train_batch_size": self.train_batch_size,
+ "predict_batch_size": self.predict_batch_size,
+ "drop_incomplete_batch": self.drop_incomplete_batch,
+ "multiproc_kwargs": self.multiproc_kwargs,
+ }
+
+ mc = get_dataset_from_module(mc_datamodule, stage, **module_args)
+ ch = get_dataset_from_module(ch_datamodule, stage, **module_args)
+ indian = get_dataset_from_module(
+ indian_datamodule, stage, **module_args
+ )
+ tbx11kv2 = get_dataset_from_module(
+ tbx11kv2_datamodule, stage, **module_args
+ )
+
+ # Combine datasets
+ self.dataset = {}
+ self.dataset["__train__"] = ConcatDataset(
+ [
+ mc["__train__"],
+ ch["__train__"],
+ indian["__train__"],
+ tbx11kv2["__train__"],
+ ]
+ )
+ self.dataset["train"] = ConcatDataset(
+ [mc["train"], ch["train"], indian["train"], tbx11kv2["train"]]
+ )
+ self.dataset["__valid__"] = ConcatDataset(
+ [
+ mc["__valid__"],
+ ch["__valid__"],
+ indian["__valid__"],
+ tbx11kv2["__valid__"],
+ ]
+ )
+ self.dataset["validation"] = ConcatDataset(
+ [
+ mc["validation"],
+ ch["validation"],
+ indian["validation"],
+ tbx11kv2["validation"],
+ ]
+ )
+ self.dataset["test"] = ConcatDataset(
+ [mc["test"], ch["test"], indian["test"], tbx11kv2["test"]]
+ )
+
+ (
+ 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/mc_ch_in_11kv2/fold_3.py b/src/ptbench/data/mc_ch_in_11kv2/fold_3.py
new file mode 100644
index 0000000000000000000000000000000000000000..11309b1c94eb182026882f934835ed9d62917b27
--- /dev/null
+++ b/src/ptbench/data/mc_ch_in_11kv2/fold_3.py
@@ -0,0 +1,101 @@
+# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
+TBX11K-simplified datasets (cross validation fold 3)"""
+
+from clapper.logging import setup
+from torch.utils.data.dataset import ConcatDataset
+
+from .. import return_subsets
+from ..base_datamodule import BaseDataModule, get_dataset_from_module
+from ..indian.fold_3 import datamodule as indian_datamodule
+from ..montgomery.fold_3 import datamodule as mc_datamodule
+from ..shenzhen.fold_3 import datamodule as ch_datamodule
+from ..tbx11k_simplified_v2.fold_3 import datamodule as tbx11kv2_datamodule
+
+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,
+ ):
+ self.train_batch_size = train_batch_size
+ self.predict_batch_size = predict_batch_size
+ self.drop_incomplete_batch = drop_incomplete_batch
+ self.multiproc_kwargs = multiproc_kwargs
+
+ 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):
+ # Instantiate other datamodules and get their datasets
+
+ module_args = {
+ "train_batch_size": self.train_batch_size,
+ "predict_batch_size": self.predict_batch_size,
+ "drop_incomplete_batch": self.drop_incomplete_batch,
+ "multiproc_kwargs": self.multiproc_kwargs,
+ }
+
+ mc = get_dataset_from_module(mc_datamodule, stage, **module_args)
+ ch = get_dataset_from_module(ch_datamodule, stage, **module_args)
+ indian = get_dataset_from_module(
+ indian_datamodule, stage, **module_args
+ )
+ tbx11kv2 = get_dataset_from_module(
+ tbx11kv2_datamodule, stage, **module_args
+ )
+
+ # Combine datasets
+ self.dataset = {}
+ self.dataset["__train__"] = ConcatDataset(
+ [
+ mc["__train__"],
+ ch["__train__"],
+ indian["__train__"],
+ tbx11kv2["__train__"],
+ ]
+ )
+ self.dataset["train"] = ConcatDataset(
+ [mc["train"], ch["train"], indian["train"], tbx11kv2["train"]]
+ )
+ self.dataset["__valid__"] = ConcatDataset(
+ [
+ mc["__valid__"],
+ ch["__valid__"],
+ indian["__valid__"],
+ tbx11kv2["__valid__"],
+ ]
+ )
+ self.dataset["validation"] = ConcatDataset(
+ [
+ mc["validation"],
+ ch["validation"],
+ indian["validation"],
+ tbx11kv2["validation"],
+ ]
+ )
+ self.dataset["test"] = ConcatDataset(
+ [mc["test"], ch["test"], indian["test"], tbx11kv2["test"]]
+ )
+
+ (
+ 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/mc_ch_in_11kv2/fold_3_rgb.py b/src/ptbench/data/mc_ch_in_11kv2/fold_3_rgb.py
new file mode 100644
index 0000000000000000000000000000000000000000..267d128e5dee412b933b350bde8ff8d60e6f771e
--- /dev/null
+++ b/src/ptbench/data/mc_ch_in_11kv2/fold_3_rgb.py
@@ -0,0 +1,101 @@
+# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
+TBX11K-simplified datasets (cross validation fold 3, RGB)"""
+
+from clapper.logging import setup
+from torch.utils.data.dataset import ConcatDataset
+
+from .. import return_subsets
+from ..base_datamodule import BaseDataModule, get_dataset_from_module
+from ..indian.fold_3_rgb import datamodule as indian_datamodule
+from ..montgomery.fold_3_rgb import datamodule as mc_datamodule
+from ..shenzhen.fold_3_rgb import datamodule as ch_datamodule
+from ..tbx11k_simplified_v2.fold_3_rgb import datamodule as tbx11kv2_datamodule
+
+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,
+ ):
+ self.train_batch_size = train_batch_size
+ self.predict_batch_size = predict_batch_size
+ self.drop_incomplete_batch = drop_incomplete_batch
+ self.multiproc_kwargs = multiproc_kwargs
+
+ 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):
+ # Instantiate other datamodules and get their datasets
+
+ module_args = {
+ "train_batch_size": self.train_batch_size,
+ "predict_batch_size": self.predict_batch_size,
+ "drop_incomplete_batch": self.drop_incomplete_batch,
+ "multiproc_kwargs": self.multiproc_kwargs,
+ }
+
+ mc = get_dataset_from_module(mc_datamodule, stage, **module_args)
+ ch = get_dataset_from_module(ch_datamodule, stage, **module_args)
+ indian = get_dataset_from_module(
+ indian_datamodule, stage, **module_args
+ )
+ tbx11kv2 = get_dataset_from_module(
+ tbx11kv2_datamodule, stage, **module_args
+ )
+
+ # Combine datasets
+ self.dataset = {}
+ self.dataset["__train__"] = ConcatDataset(
+ [
+ mc["__train__"],
+ ch["__train__"],
+ indian["__train__"],
+ tbx11kv2["__train__"],
+ ]
+ )
+ self.dataset["train"] = ConcatDataset(
+ [mc["train"], ch["train"], indian["train"], tbx11kv2["train"]]
+ )
+ self.dataset["__valid__"] = ConcatDataset(
+ [
+ mc["__valid__"],
+ ch["__valid__"],
+ indian["__valid__"],
+ tbx11kv2["__valid__"],
+ ]
+ )
+ self.dataset["validation"] = ConcatDataset(
+ [
+ mc["validation"],
+ ch["validation"],
+ indian["validation"],
+ tbx11kv2["validation"],
+ ]
+ )
+ self.dataset["test"] = ConcatDataset(
+ [mc["test"], ch["test"], indian["test"], tbx11kv2["test"]]
+ )
+
+ (
+ 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/mc_ch_in_11kv2/fold_4.py b/src/ptbench/data/mc_ch_in_11kv2/fold_4.py
new file mode 100644
index 0000000000000000000000000000000000000000..0f53ed635e2690776ca704bc50f812cc09c33256
--- /dev/null
+++ b/src/ptbench/data/mc_ch_in_11kv2/fold_4.py
@@ -0,0 +1,101 @@
+# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
+TBX11K-simplified datasets (cross validation fold 4)"""
+
+from clapper.logging import setup
+from torch.utils.data.dataset import ConcatDataset
+
+from .. import return_subsets
+from ..base_datamodule import BaseDataModule, get_dataset_from_module
+from ..indian.fold_4 import datamodule as indian_datamodule
+from ..montgomery.fold_4 import datamodule as mc_datamodule
+from ..shenzhen.fold_4 import datamodule as ch_datamodule
+from ..tbx11k_simplified_v2.fold_4 import datamodule as tbx11kv2_datamodule
+
+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,
+ ):
+ self.train_batch_size = train_batch_size
+ self.predict_batch_size = predict_batch_size
+ self.drop_incomplete_batch = drop_incomplete_batch
+ self.multiproc_kwargs = multiproc_kwargs
+
+ 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):
+ # Instantiate other datamodules and get their datasets
+
+ module_args = {
+ "train_batch_size": self.train_batch_size,
+ "predict_batch_size": self.predict_batch_size,
+ "drop_incomplete_batch": self.drop_incomplete_batch,
+ "multiproc_kwargs": self.multiproc_kwargs,
+ }
+
+ mc = get_dataset_from_module(mc_datamodule, stage, **module_args)
+ ch = get_dataset_from_module(ch_datamodule, stage, **module_args)
+ indian = get_dataset_from_module(
+ indian_datamodule, stage, **module_args
+ )
+ tbx11kv2 = get_dataset_from_module(
+ tbx11kv2_datamodule, stage, **module_args
+ )
+
+ # Combine datasets
+ self.dataset = {}
+ self.dataset["__train__"] = ConcatDataset(
+ [
+ mc["__train__"],
+ ch["__train__"],
+ indian["__train__"],
+ tbx11kv2["__train__"],
+ ]
+ )
+ self.dataset["train"] = ConcatDataset(
+ [mc["train"], ch["train"], indian["train"], tbx11kv2["train"]]
+ )
+ self.dataset["__valid__"] = ConcatDataset(
+ [
+ mc["__valid__"],
+ ch["__valid__"],
+ indian["__valid__"],
+ tbx11kv2["__valid__"],
+ ]
+ )
+ self.dataset["validation"] = ConcatDataset(
+ [
+ mc["validation"],
+ ch["validation"],
+ indian["validation"],
+ tbx11kv2["validation"],
+ ]
+ )
+ self.dataset["test"] = ConcatDataset(
+ [mc["test"], ch["test"], indian["test"], tbx11kv2["test"]]
+ )
+
+ (
+ 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/mc_ch_in_11kv2/fold_4_rgb.py b/src/ptbench/data/mc_ch_in_11kv2/fold_4_rgb.py
new file mode 100644
index 0000000000000000000000000000000000000000..9bdc8c23043961b780f8b53d4349ce9cd4c7e3fd
--- /dev/null
+++ b/src/ptbench/data/mc_ch_in_11kv2/fold_4_rgb.py
@@ -0,0 +1,101 @@
+# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
+TBX11K-simplified datasets (cross validation fold 4, RGB)"""
+
+from clapper.logging import setup
+from torch.utils.data.dataset import ConcatDataset
+
+from .. import return_subsets
+from ..base_datamodule import BaseDataModule, get_dataset_from_module
+from ..indian.fold_4_rgb import datamodule as indian_datamodule
+from ..montgomery.fold_4_rgb import datamodule as mc_datamodule
+from ..shenzhen.fold_4_rgb import datamodule as ch_datamodule
+from ..tbx11k_simplified_v2.fold_4_rgb import datamodule as tbx11kv2_datamodule
+
+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,
+ ):
+ self.train_batch_size = train_batch_size
+ self.predict_batch_size = predict_batch_size
+ self.drop_incomplete_batch = drop_incomplete_batch
+ self.multiproc_kwargs = multiproc_kwargs
+
+ 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):
+ # Instantiate other datamodules and get their datasets
+
+ module_args = {
+ "train_batch_size": self.train_batch_size,
+ "predict_batch_size": self.predict_batch_size,
+ "drop_incomplete_batch": self.drop_incomplete_batch,
+ "multiproc_kwargs": self.multiproc_kwargs,
+ }
+
+ mc = get_dataset_from_module(mc_datamodule, stage, **module_args)
+ ch = get_dataset_from_module(ch_datamodule, stage, **module_args)
+ indian = get_dataset_from_module(
+ indian_datamodule, stage, **module_args
+ )
+ tbx11kv2 = get_dataset_from_module(
+ tbx11kv2_datamodule, stage, **module_args
+ )
+
+ # Combine datasets
+ self.dataset = {}
+ self.dataset["__train__"] = ConcatDataset(
+ [
+ mc["__train__"],
+ ch["__train__"],
+ indian["__train__"],
+ tbx11kv2["__train__"],
+ ]
+ )
+ self.dataset["train"] = ConcatDataset(
+ [mc["train"], ch["train"], indian["train"], tbx11kv2["train"]]
+ )
+ self.dataset["__valid__"] = ConcatDataset(
+ [
+ mc["__valid__"],
+ ch["__valid__"],
+ indian["__valid__"],
+ tbx11kv2["__valid__"],
+ ]
+ )
+ self.dataset["validation"] = ConcatDataset(
+ [
+ mc["validation"],
+ ch["validation"],
+ indian["validation"],
+ tbx11kv2["validation"],
+ ]
+ )
+ self.dataset["test"] = ConcatDataset(
+ [mc["test"], ch["test"], indian["test"], tbx11kv2["test"]]
+ )
+
+ (
+ 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/mc_ch_in_11kv2/fold_5.py b/src/ptbench/data/mc_ch_in_11kv2/fold_5.py
new file mode 100644
index 0000000000000000000000000000000000000000..df7dc0e8fe8fd9a6ce122a20161afdb927aa62eb
--- /dev/null
+++ b/src/ptbench/data/mc_ch_in_11kv2/fold_5.py
@@ -0,0 +1,101 @@
+# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
+TBX11K-simplified datasets (cross validation fold 5)"""
+
+from clapper.logging import setup
+from torch.utils.data.dataset import ConcatDataset
+
+from .. import return_subsets
+from ..base_datamodule import BaseDataModule, get_dataset_from_module
+from ..indian.fold_5 import datamodule as indian_datamodule
+from ..montgomery.fold_5 import datamodule as mc_datamodule
+from ..shenzhen.fold_5 import datamodule as ch_datamodule
+from ..tbx11k_simplified_v2.fold_5 import datamodule as tbx11kv2_datamodule
+
+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,
+ ):
+ self.train_batch_size = train_batch_size
+ self.predict_batch_size = predict_batch_size
+ self.drop_incomplete_batch = drop_incomplete_batch
+ self.multiproc_kwargs = multiproc_kwargs
+
+ 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):
+ # Instantiate other datamodules and get their datasets
+
+ module_args = {
+ "train_batch_size": self.train_batch_size,
+ "predict_batch_size": self.predict_batch_size,
+ "drop_incomplete_batch": self.drop_incomplete_batch,
+ "multiproc_kwargs": self.multiproc_kwargs,
+ }
+
+ mc = get_dataset_from_module(mc_datamodule, stage, **module_args)
+ ch = get_dataset_from_module(ch_datamodule, stage, **module_args)
+ indian = get_dataset_from_module(
+ indian_datamodule, stage, **module_args
+ )
+ tbx11kv2 = get_dataset_from_module(
+ tbx11kv2_datamodule, stage, **module_args
+ )
+
+ # Combine datasets
+ self.dataset = {}
+ self.dataset["__train__"] = ConcatDataset(
+ [
+ mc["__train__"],
+ ch["__train__"],
+ indian["__train__"],
+ tbx11kv2["__train__"],
+ ]
+ )
+ self.dataset["train"] = ConcatDataset(
+ [mc["train"], ch["train"], indian["train"], tbx11kv2["train"]]
+ )
+ self.dataset["__valid__"] = ConcatDataset(
+ [
+ mc["__valid__"],
+ ch["__valid__"],
+ indian["__valid__"],
+ tbx11kv2["__valid__"],
+ ]
+ )
+ self.dataset["validation"] = ConcatDataset(
+ [
+ mc["validation"],
+ ch["validation"],
+ indian["validation"],
+ tbx11kv2["validation"],
+ ]
+ )
+ self.dataset["test"] = ConcatDataset(
+ [mc["test"], ch["test"], indian["test"], tbx11kv2["test"]]
+ )
+
+ (
+ 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/mc_ch_in_11kv2/fold_5_rgb.py b/src/ptbench/data/mc_ch_in_11kv2/fold_5_rgb.py
new file mode 100644
index 0000000000000000000000000000000000000000..ce3ffd479d9264bd17188e89cfc8716f59caf25a
--- /dev/null
+++ b/src/ptbench/data/mc_ch_in_11kv2/fold_5_rgb.py
@@ -0,0 +1,101 @@
+# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
+TBX11K-simplified datasets (cross validation fold 5, RGB)"""
+
+from clapper.logging import setup
+from torch.utils.data.dataset import ConcatDataset
+
+from .. import return_subsets
+from ..base_datamodule import BaseDataModule, get_dataset_from_module
+from ..indian.fold_5_rgb import datamodule as indian_datamodule
+from ..montgomery.fold_5_rgb import datamodule as mc_datamodule
+from ..shenzhen.fold_5_rgb import datamodule as ch_datamodule
+from ..tbx11k_simplified_v2.fold_5_rgb import datamodule as tbx11kv2_datamodule
+
+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,
+ ):
+ self.train_batch_size = train_batch_size
+ self.predict_batch_size = predict_batch_size
+ self.drop_incomplete_batch = drop_incomplete_batch
+ self.multiproc_kwargs = multiproc_kwargs
+
+ 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):
+ # Instantiate other datamodules and get their datasets
+
+ module_args = {
+ "train_batch_size": self.train_batch_size,
+ "predict_batch_size": self.predict_batch_size,
+ "drop_incomplete_batch": self.drop_incomplete_batch,
+ "multiproc_kwargs": self.multiproc_kwargs,
+ }
+
+ mc = get_dataset_from_module(mc_datamodule, stage, **module_args)
+ ch = get_dataset_from_module(ch_datamodule, stage, **module_args)
+ indian = get_dataset_from_module(
+ indian_datamodule, stage, **module_args
+ )
+ tbx11kv2 = get_dataset_from_module(
+ tbx11kv2_datamodule, stage, **module_args
+ )
+
+ # Combine datasets
+ self.dataset = {}
+ self.dataset["__train__"] = ConcatDataset(
+ [
+ mc["__train__"],
+ ch["__train__"],
+ indian["__train__"],
+ tbx11kv2["__train__"],
+ ]
+ )
+ self.dataset["train"] = ConcatDataset(
+ [mc["train"], ch["train"], indian["train"], tbx11kv2["train"]]
+ )
+ self.dataset["__valid__"] = ConcatDataset(
+ [
+ mc["__valid__"],
+ ch["__valid__"],
+ indian["__valid__"],
+ tbx11kv2["__valid__"],
+ ]
+ )
+ self.dataset["validation"] = ConcatDataset(
+ [
+ mc["validation"],
+ ch["validation"],
+ indian["validation"],
+ tbx11kv2["validation"],
+ ]
+ )
+ self.dataset["test"] = ConcatDataset(
+ [mc["test"], ch["test"], indian["test"], tbx11kv2["test"]]
+ )
+
+ (
+ 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/mc_ch_in_11kv2/fold_6.py b/src/ptbench/data/mc_ch_in_11kv2/fold_6.py
new file mode 100644
index 0000000000000000000000000000000000000000..ef246cd82afe20f65d1d214646155fe9416be479
--- /dev/null
+++ b/src/ptbench/data/mc_ch_in_11kv2/fold_6.py
@@ -0,0 +1,101 @@
+# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
+TBX11K-simplified datasets (cross validation fold 6)"""
+
+from clapper.logging import setup
+from torch.utils.data.dataset import ConcatDataset
+
+from .. import return_subsets
+from ..base_datamodule import BaseDataModule, get_dataset_from_module
+from ..indian.fold_6 import datamodule as indian_datamodule
+from ..montgomery.fold_6 import datamodule as mc_datamodule
+from ..shenzhen.fold_6 import datamodule as ch_datamodule
+from ..tbx11k_simplified_v2.fold_6 import datamodule as tbx11kv2_datamodule
+
+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,
+ ):
+ self.train_batch_size = train_batch_size
+ self.predict_batch_size = predict_batch_size
+ self.drop_incomplete_batch = drop_incomplete_batch
+ self.multiproc_kwargs = multiproc_kwargs
+
+ 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):
+ # Instantiate other datamodules and get their datasets
+
+ module_args = {
+ "train_batch_size": self.train_batch_size,
+ "predict_batch_size": self.predict_batch_size,
+ "drop_incomplete_batch": self.drop_incomplete_batch,
+ "multiproc_kwargs": self.multiproc_kwargs,
+ }
+
+ mc = get_dataset_from_module(mc_datamodule, stage, **module_args)
+ ch = get_dataset_from_module(ch_datamodule, stage, **module_args)
+ indian = get_dataset_from_module(
+ indian_datamodule, stage, **module_args
+ )
+ tbx11kv2 = get_dataset_from_module(
+ tbx11kv2_datamodule, stage, **module_args
+ )
+
+ # Combine datasets
+ self.dataset = {}
+ self.dataset["__train__"] = ConcatDataset(
+ [
+ mc["__train__"],
+ ch["__train__"],
+ indian["__train__"],
+ tbx11kv2["__train__"],
+ ]
+ )
+ self.dataset["train"] = ConcatDataset(
+ [mc["train"], ch["train"], indian["train"], tbx11kv2["train"]]
+ )
+ self.dataset["__valid__"] = ConcatDataset(
+ [
+ mc["__valid__"],
+ ch["__valid__"],
+ indian["__valid__"],
+ tbx11kv2["__valid__"],
+ ]
+ )
+ self.dataset["validation"] = ConcatDataset(
+ [
+ mc["validation"],
+ ch["validation"],
+ indian["validation"],
+ tbx11kv2["validation"],
+ ]
+ )
+ self.dataset["test"] = ConcatDataset(
+ [mc["test"], ch["test"], indian["test"], tbx11kv2["test"]]
+ )
+
+ (
+ 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/mc_ch_in_11kv2/fold_6_rgb.py b/src/ptbench/data/mc_ch_in_11kv2/fold_6_rgb.py
new file mode 100644
index 0000000000000000000000000000000000000000..bc892b7964c8565d70577b514be426be85073645
--- /dev/null
+++ b/src/ptbench/data/mc_ch_in_11kv2/fold_6_rgb.py
@@ -0,0 +1,101 @@
+# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
+TBX11K-simplified datasets (cross validation fold 6, RGB)"""
+
+from clapper.logging import setup
+from torch.utils.data.dataset import ConcatDataset
+
+from .. import return_subsets
+from ..base_datamodule import BaseDataModule, get_dataset_from_module
+from ..indian.fold_6_rgb import datamodule as indian_datamodule
+from ..montgomery.fold_6_rgb import datamodule as mc_datamodule
+from ..shenzhen.fold_5_rgb import datamodule as ch_datamodule
+from ..tbx11k_simplified_v2.fold_6_rgb import datamodule as tbx11kv2_datamodule
+
+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,
+ ):
+ self.train_batch_size = train_batch_size
+ self.predict_batch_size = predict_batch_size
+ self.drop_incomplete_batch = drop_incomplete_batch
+ self.multiproc_kwargs = multiproc_kwargs
+
+ 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):
+ # Instantiate other datamodules and get their datasets
+
+ module_args = {
+ "train_batch_size": self.train_batch_size,
+ "predict_batch_size": self.predict_batch_size,
+ "drop_incomplete_batch": self.drop_incomplete_batch,
+ "multiproc_kwargs": self.multiproc_kwargs,
+ }
+
+ mc = get_dataset_from_module(mc_datamodule, stage, **module_args)
+ ch = get_dataset_from_module(ch_datamodule, stage, **module_args)
+ indian = get_dataset_from_module(
+ indian_datamodule, stage, **module_args
+ )
+ tbx11kv2 = get_dataset_from_module(
+ tbx11kv2_datamodule, stage, **module_args
+ )
+
+ # Combine datasets
+ self.dataset = {}
+ self.dataset["__train__"] = ConcatDataset(
+ [
+ mc["__train__"],
+ ch["__train__"],
+ indian["__train__"],
+ tbx11kv2["__train__"],
+ ]
+ )
+ self.dataset["train"] = ConcatDataset(
+ [mc["train"], ch["train"], indian["train"], tbx11kv2["train"]]
+ )
+ self.dataset["__valid__"] = ConcatDataset(
+ [
+ mc["__valid__"],
+ ch["__valid__"],
+ indian["__valid__"],
+ tbx11kv2["__valid__"],
+ ]
+ )
+ self.dataset["validation"] = ConcatDataset(
+ [
+ mc["validation"],
+ ch["validation"],
+ indian["validation"],
+ tbx11kv2["validation"],
+ ]
+ )
+ self.dataset["test"] = ConcatDataset(
+ [mc["test"], ch["test"], indian["test"], tbx11kv2["test"]]
+ )
+
+ (
+ 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/mc_ch_in_11kv2/fold_7.py b/src/ptbench/data/mc_ch_in_11kv2/fold_7.py
new file mode 100644
index 0000000000000000000000000000000000000000..41d3ba3ebf059320e36195d85447fe57bf1dffbb
--- /dev/null
+++ b/src/ptbench/data/mc_ch_in_11kv2/fold_7.py
@@ -0,0 +1,101 @@
+# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
+TBX11K-simplified datasets (cross validation fold 7)"""
+
+from clapper.logging import setup
+from torch.utils.data.dataset import ConcatDataset
+
+from .. import return_subsets
+from ..base_datamodule import BaseDataModule, get_dataset_from_module
+from ..indian.fold_7 import datamodule as indian_datamodule
+from ..montgomery.fold_7 import datamodule as mc_datamodule
+from ..shenzhen.fold_7 import datamodule as ch_datamodule
+from ..tbx11k_simplified_v2.fold_7 import datamodule as tbx11kv2_datamodule
+
+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,
+ ):
+ self.train_batch_size = train_batch_size
+ self.predict_batch_size = predict_batch_size
+ self.drop_incomplete_batch = drop_incomplete_batch
+ self.multiproc_kwargs = multiproc_kwargs
+
+ 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):
+ # Instantiate other datamodules and get their datasets
+
+ module_args = {
+ "train_batch_size": self.train_batch_size,
+ "predict_batch_size": self.predict_batch_size,
+ "drop_incomplete_batch": self.drop_incomplete_batch,
+ "multiproc_kwargs": self.multiproc_kwargs,
+ }
+
+ mc = get_dataset_from_module(mc_datamodule, stage, **module_args)
+ ch = get_dataset_from_module(ch_datamodule, stage, **module_args)
+ indian = get_dataset_from_module(
+ indian_datamodule, stage, **module_args
+ )
+ tbx11kv2 = get_dataset_from_module(
+ tbx11kv2_datamodule, stage, **module_args
+ )
+
+ # Combine datasets
+ self.dataset = {}
+ self.dataset["__train__"] = ConcatDataset(
+ [
+ mc["__train__"],
+ ch["__train__"],
+ indian["__train__"],
+ tbx11kv2["__train__"],
+ ]
+ )
+ self.dataset["train"] = ConcatDataset(
+ [mc["train"], ch["train"], indian["train"], tbx11kv2["train"]]
+ )
+ self.dataset["__valid__"] = ConcatDataset(
+ [
+ mc["__valid__"],
+ ch["__valid__"],
+ indian["__valid__"],
+ tbx11kv2["__valid__"],
+ ]
+ )
+ self.dataset["validation"] = ConcatDataset(
+ [
+ mc["validation"],
+ ch["validation"],
+ indian["validation"],
+ tbx11kv2["validation"],
+ ]
+ )
+ self.dataset["test"] = ConcatDataset(
+ [mc["test"], ch["test"], indian["test"], tbx11kv2["test"]]
+ )
+
+ (
+ 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/mc_ch_in_11kv2/fold_7_rgb.py b/src/ptbench/data/mc_ch_in_11kv2/fold_7_rgb.py
new file mode 100644
index 0000000000000000000000000000000000000000..0fe1f28289a5a09f5fd321de5d8cd65c57fef4e0
--- /dev/null
+++ b/src/ptbench/data/mc_ch_in_11kv2/fold_7_rgb.py
@@ -0,0 +1,101 @@
+# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
+TBX11K-simplified datasets (cross validation fold 7, RGB)"""
+
+from clapper.logging import setup
+from torch.utils.data.dataset import ConcatDataset
+
+from .. import return_subsets
+from ..base_datamodule import BaseDataModule, get_dataset_from_module
+from ..indian.fold_7_rgb import datamodule as indian_datamodule
+from ..montgomery.fold_7_rgb import datamodule as mc_datamodule
+from ..shenzhen.fold_7_rgb import datamodule as ch_datamodule
+from ..tbx11k_simplified_v2.fold_7_rgb import datamodule as tbx11kv2_datamodule
+
+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,
+ ):
+ self.train_batch_size = train_batch_size
+ self.predict_batch_size = predict_batch_size
+ self.drop_incomplete_batch = drop_incomplete_batch
+ self.multiproc_kwargs = multiproc_kwargs
+
+ 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):
+ # Instantiate other datamodules and get their datasets
+
+ module_args = {
+ "train_batch_size": self.train_batch_size,
+ "predict_batch_size": self.predict_batch_size,
+ "drop_incomplete_batch": self.drop_incomplete_batch,
+ "multiproc_kwargs": self.multiproc_kwargs,
+ }
+
+ mc = get_dataset_from_module(mc_datamodule, stage, **module_args)
+ ch = get_dataset_from_module(ch_datamodule, stage, **module_args)
+ indian = get_dataset_from_module(
+ indian_datamodule, stage, **module_args
+ )
+ tbx11kv2 = get_dataset_from_module(
+ tbx11kv2_datamodule, stage, **module_args
+ )
+
+ # Combine datasets
+ self.dataset = {}
+ self.dataset["__train__"] = ConcatDataset(
+ [
+ mc["__train__"],
+ ch["__train__"],
+ indian["__train__"],
+ tbx11kv2["__train__"],
+ ]
+ )
+ self.dataset["train"] = ConcatDataset(
+ [mc["train"], ch["train"], indian["train"], tbx11kv2["train"]]
+ )
+ self.dataset["__valid__"] = ConcatDataset(
+ [
+ mc["__valid__"],
+ ch["__valid__"],
+ indian["__valid__"],
+ tbx11kv2["__valid__"],
+ ]
+ )
+ self.dataset["validation"] = ConcatDataset(
+ [
+ mc["validation"],
+ ch["validation"],
+ indian["validation"],
+ tbx11kv2["validation"],
+ ]
+ )
+ self.dataset["test"] = ConcatDataset(
+ [mc["test"], ch["test"], indian["test"], tbx11kv2["test"]]
+ )
+
+ (
+ 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/mc_ch_in_11kv2/fold_8.py b/src/ptbench/data/mc_ch_in_11kv2/fold_8.py
new file mode 100644
index 0000000000000000000000000000000000000000..a6842d47c932586d6910e05df039921fc82c5843
--- /dev/null
+++ b/src/ptbench/data/mc_ch_in_11kv2/fold_8.py
@@ -0,0 +1,101 @@
+# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
+TBX11K-simplified datasets (cross validation fold 8)"""
+
+from clapper.logging import setup
+from torch.utils.data.dataset import ConcatDataset
+
+from .. import return_subsets
+from ..base_datamodule import BaseDataModule, get_dataset_from_module
+from ..indian.fold_8 import datamodule as indian_datamodule
+from ..montgomery.fold_8 import datamodule as mc_datamodule
+from ..shenzhen.fold_8 import datamodule as ch_datamodule
+from ..tbx11k_simplified_v2.fold_8 import datamodule as tbx11kv2_datamodule
+
+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,
+ ):
+ self.train_batch_size = train_batch_size
+ self.predict_batch_size = predict_batch_size
+ self.drop_incomplete_batch = drop_incomplete_batch
+ self.multiproc_kwargs = multiproc_kwargs
+
+ 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):
+ # Instantiate other datamodules and get their datasets
+
+ module_args = {
+ "train_batch_size": self.train_batch_size,
+ "predict_batch_size": self.predict_batch_size,
+ "drop_incomplete_batch": self.drop_incomplete_batch,
+ "multiproc_kwargs": self.multiproc_kwargs,
+ }
+
+ mc = get_dataset_from_module(mc_datamodule, stage, **module_args)
+ ch = get_dataset_from_module(ch_datamodule, stage, **module_args)
+ indian = get_dataset_from_module(
+ indian_datamodule, stage, **module_args
+ )
+ tbx11kv2 = get_dataset_from_module(
+ tbx11kv2_datamodule, stage, **module_args
+ )
+
+ # Combine datasets
+ self.dataset = {}
+ self.dataset["__train__"] = ConcatDataset(
+ [
+ mc["__train__"],
+ ch["__train__"],
+ indian["__train__"],
+ tbx11kv2["__train__"],
+ ]
+ )
+ self.dataset["train"] = ConcatDataset(
+ [mc["train"], ch["train"], indian["train"], tbx11kv2["train"]]
+ )
+ self.dataset["__valid__"] = ConcatDataset(
+ [
+ mc["__valid__"],
+ ch["__valid__"],
+ indian["__valid__"],
+ tbx11kv2["__valid__"],
+ ]
+ )
+ self.dataset["validation"] = ConcatDataset(
+ [
+ mc["validation"],
+ ch["validation"],
+ indian["validation"],
+ tbx11kv2["validation"],
+ ]
+ )
+ self.dataset["test"] = ConcatDataset(
+ [mc["test"], ch["test"], indian["test"], tbx11kv2["test"]]
+ )
+
+ (
+ 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/mc_ch_in_11kv2/fold_8_rgb.py b/src/ptbench/data/mc_ch_in_11kv2/fold_8_rgb.py
new file mode 100644
index 0000000000000000000000000000000000000000..a330b2e6d12847bc6406d87e5d48ad626f802c77
--- /dev/null
+++ b/src/ptbench/data/mc_ch_in_11kv2/fold_8_rgb.py
@@ -0,0 +1,101 @@
+# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
+TBX11K-simplified datasets (cross validation fold 8, RGB)"""
+
+from clapper.logging import setup
+from torch.utils.data.dataset import ConcatDataset
+
+from .. import return_subsets
+from ..base_datamodule import BaseDataModule, get_dataset_from_module
+from ..indian.fold_8_rgb import datamodule as indian_datamodule
+from ..montgomery.fold_8_rgb import datamodule as mc_datamodule
+from ..shenzhen.fold_8_rgb import datamodule as ch_datamodule
+from ..tbx11k_simplified_v2.fold_8_rgb import datamodule as tbx11kv2_datamodule
+
+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,
+ ):
+ self.train_batch_size = train_batch_size
+ self.predict_batch_size = predict_batch_size
+ self.drop_incomplete_batch = drop_incomplete_batch
+ self.multiproc_kwargs = multiproc_kwargs
+
+ 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):
+ # Instantiate other datamodules and get their datasets
+
+ module_args = {
+ "train_batch_size": self.train_batch_size,
+ "predict_batch_size": self.predict_batch_size,
+ "drop_incomplete_batch": self.drop_incomplete_batch,
+ "multiproc_kwargs": self.multiproc_kwargs,
+ }
+
+ mc = get_dataset_from_module(mc_datamodule, stage, **module_args)
+ ch = get_dataset_from_module(ch_datamodule, stage, **module_args)
+ indian = get_dataset_from_module(
+ indian_datamodule, stage, **module_args
+ )
+ tbx11kv2 = get_dataset_from_module(
+ tbx11kv2_datamodule, stage, **module_args
+ )
+
+ # Combine datasets
+ self.dataset = {}
+ self.dataset["__train__"] = ConcatDataset(
+ [
+ mc["__train__"],
+ ch["__train__"],
+ indian["__train__"],
+ tbx11kv2["__train__"],
+ ]
+ )
+ self.dataset["train"] = ConcatDataset(
+ [mc["train"], ch["train"], indian["train"], tbx11kv2["train"]]
+ )
+ self.dataset["__valid__"] = ConcatDataset(
+ [
+ mc["__valid__"],
+ ch["__valid__"],
+ indian["__valid__"],
+ tbx11kv2["__valid__"],
+ ]
+ )
+ self.dataset["validation"] = ConcatDataset(
+ [
+ mc["validation"],
+ ch["validation"],
+ indian["validation"],
+ tbx11kv2["validation"],
+ ]
+ )
+ self.dataset["test"] = ConcatDataset(
+ [mc["test"], ch["test"], indian["test"], tbx11kv2["test"]]
+ )
+
+ (
+ 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/mc_ch_in_11kv2/fold_9.py b/src/ptbench/data/mc_ch_in_11kv2/fold_9.py
new file mode 100644
index 0000000000000000000000000000000000000000..ba69b78892058b141a761e6cb4070683a0098fd8
--- /dev/null
+++ b/src/ptbench/data/mc_ch_in_11kv2/fold_9.py
@@ -0,0 +1,101 @@
+# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
+TBX11K-simplified datasets (cross validation fold 9)"""
+
+from clapper.logging import setup
+from torch.utils.data.dataset import ConcatDataset
+
+from .. import return_subsets
+from ..base_datamodule import BaseDataModule, get_dataset_from_module
+from ..indian.fold_9 import datamodule as indian_datamodule
+from ..montgomery.fold_9 import datamodule as mc_datamodule
+from ..shenzhen.fold_9 import datamodule as ch_datamodule
+from ..tbx11k_simplified_v2.fold_9 import datamodule as tbx11kv2_datamodule
+
+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,
+ ):
+ self.train_batch_size = train_batch_size
+ self.predict_batch_size = predict_batch_size
+ self.drop_incomplete_batch = drop_incomplete_batch
+ self.multiproc_kwargs = multiproc_kwargs
+
+ 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):
+ # Instantiate other datamodules and get their datasets
+
+ module_args = {
+ "train_batch_size": self.train_batch_size,
+ "predict_batch_size": self.predict_batch_size,
+ "drop_incomplete_batch": self.drop_incomplete_batch,
+ "multiproc_kwargs": self.multiproc_kwargs,
+ }
+
+ mc = get_dataset_from_module(mc_datamodule, stage, **module_args)
+ ch = get_dataset_from_module(ch_datamodule, stage, **module_args)
+ indian = get_dataset_from_module(
+ indian_datamodule, stage, **module_args
+ )
+ tbx11kv2 = get_dataset_from_module(
+ tbx11kv2_datamodule, stage, **module_args
+ )
+
+ # Combine datasets
+ self.dataset = {}
+ self.dataset["__train__"] = ConcatDataset(
+ [
+ mc["__train__"],
+ ch["__train__"],
+ indian["__train__"],
+ tbx11kv2["__train__"],
+ ]
+ )
+ self.dataset["train"] = ConcatDataset(
+ [mc["train"], ch["train"], indian["train"], tbx11kv2["train"]]
+ )
+ self.dataset["__valid__"] = ConcatDataset(
+ [
+ mc["__valid__"],
+ ch["__valid__"],
+ indian["__valid__"],
+ tbx11kv2["__valid__"],
+ ]
+ )
+ self.dataset["validation"] = ConcatDataset(
+ [
+ mc["validation"],
+ ch["validation"],
+ indian["validation"],
+ tbx11kv2["validation"],
+ ]
+ )
+ self.dataset["test"] = ConcatDataset(
+ [mc["test"], ch["test"], indian["test"], tbx11kv2["test"]]
+ )
+
+ (
+ 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/mc_ch_in_11kv2/fold_9_rgb.py b/src/ptbench/data/mc_ch_in_11kv2/fold_9_rgb.py
new file mode 100644
index 0000000000000000000000000000000000000000..93296414df6d566de6a48556074b185c045cac08
--- /dev/null
+++ b/src/ptbench/data/mc_ch_in_11kv2/fold_9_rgb.py
@@ -0,0 +1,101 @@
+# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
+TBX11K-simplified datasets (cross validation fold 9, RGB)"""
+
+from clapper.logging import setup
+from torch.utils.data.dataset import ConcatDataset
+
+from .. import return_subsets
+from ..base_datamodule import BaseDataModule, get_dataset_from_module
+from ..indian.fold_9_rgb import datamodule as indian_datamodule
+from ..montgomery.fold_9_rgb import datamodule as mc_datamodule
+from ..shenzhen.fold_9_rgb import datamodule as ch_datamodule
+from ..tbx11k_simplified_v2.fold_9_rgb import datamodule as tbx11kv2_datamodule
+
+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,
+ ):
+ self.train_batch_size = train_batch_size
+ self.predict_batch_size = predict_batch_size
+ self.drop_incomplete_batch = drop_incomplete_batch
+ self.multiproc_kwargs = multiproc_kwargs
+
+ 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):
+ # Instantiate other datamodules and get their datasets
+
+ module_args = {
+ "train_batch_size": self.train_batch_size,
+ "predict_batch_size": self.predict_batch_size,
+ "drop_incomplete_batch": self.drop_incomplete_batch,
+ "multiproc_kwargs": self.multiproc_kwargs,
+ }
+
+ mc = get_dataset_from_module(mc_datamodule, stage, **module_args)
+ ch = get_dataset_from_module(ch_datamodule, stage, **module_args)
+ indian = get_dataset_from_module(
+ indian_datamodule, stage, **module_args
+ )
+ tbx11kv2 = get_dataset_from_module(
+ tbx11kv2_datamodule, stage, **module_args
+ )
+
+ # Combine datasets
+ self.dataset = {}
+ self.dataset["__train__"] = ConcatDataset(
+ [
+ mc["__train__"],
+ ch["__train__"],
+ indian["__train__"],
+ tbx11kv2["__train__"],
+ ]
+ )
+ self.dataset["train"] = ConcatDataset(
+ [mc["train"], ch["train"], indian["train"], tbx11kv2["train"]]
+ )
+ self.dataset["__valid__"] = ConcatDataset(
+ [
+ mc["__valid__"],
+ ch["__valid__"],
+ indian["__valid__"],
+ tbx11kv2["__valid__"],
+ ]
+ )
+ self.dataset["validation"] = ConcatDataset(
+ [
+ mc["validation"],
+ ch["validation"],
+ indian["validation"],
+ tbx11kv2["validation"],
+ ]
+ )
+ self.dataset["test"] = ConcatDataset(
+ [mc["test"], ch["test"], indian["test"], tbx11kv2["test"]]
+ )
+
+ (
+ 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/mc_ch_in_11kv2/rgb.py b/src/ptbench/data/mc_ch_in_11kv2/rgb.py
new file mode 100644
index 0000000000000000000000000000000000000000..c99b8fe902afb572dd3dae0dd2766b8fd9d961cb
--- /dev/null
+++ b/src/ptbench/data/mc_ch_in_11kv2/rgb.py
@@ -0,0 +1,101 @@
+# Copyright © 2022 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+"""Aggregated dataset composed of Montgomery, Shenzhen, Indian and the default
+TBX11K-simplified datasets (RGB)"""
+
+from clapper.logging import setup
+from torch.utils.data.dataset import ConcatDataset
+
+from .. import return_subsets
+from ..base_datamodule import BaseDataModule, get_dataset_from_module
+from ..indian.rgb import datamodule as indian_datamodule
+from ..montgomery.rgb import datamodule as mc_datamodule
+from ..shenzhen.rgb import datamodule as ch_datamodule
+from ..tbx11k_simplified_v2.rgb import datamodule as tbx11kv2_datamodule
+
+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,
+ ):
+ self.train_batch_size = train_batch_size
+ self.predict_batch_size = predict_batch_size
+ self.drop_incomplete_batch = drop_incomplete_batch
+ self.multiproc_kwargs = multiproc_kwargs
+
+ 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):
+ # Instantiate other datamodules and get their datasets
+
+ module_args = {
+ "train_batch_size": self.train_batch_size,
+ "predict_batch_size": self.predict_batch_size,
+ "drop_incomplete_batch": self.drop_incomplete_batch,
+ "multiproc_kwargs": self.multiproc_kwargs,
+ }
+
+ mc = get_dataset_from_module(mc_datamodule, stage, **module_args)
+ ch = get_dataset_from_module(ch_datamodule, stage, **module_args)
+ indian = get_dataset_from_module(
+ indian_datamodule, stage, **module_args
+ )
+ tbx11kv2 = get_dataset_from_module(
+ tbx11kv2_datamodule, stage, **module_args
+ )
+
+ # Combine datasets
+ self.dataset = {}
+ self.dataset["__train__"] = ConcatDataset(
+ [
+ mc["__train__"],
+ ch["__train__"],
+ indian["__train__"],
+ tbx11kv2["__train__"],
+ ]
+ )
+ self.dataset["train"] = ConcatDataset(
+ [mc["train"], ch["train"], indian["train"], tbx11kv2["train"]]
+ )
+ self.dataset["__valid__"] = ConcatDataset(
+ [
+ mc["__valid__"],
+ ch["__valid__"],
+ indian["__valid__"],
+ tbx11kv2["__valid__"],
+ ]
+ )
+ self.dataset["validation"] = ConcatDataset(
+ [
+ mc["validation"],
+ ch["validation"],
+ indian["validation"],
+ tbx11kv2["validation"],
+ ]
+ )
+ self.dataset["test"] = ConcatDataset(
+ [mc["test"], ch["test"], indian["test"], tbx11kv2["test"]]
+ )
+
+ (
+ self.train_dataset,
+ self.validation_dataset,
+ self.extra_validation_datasets,
+ self.predict_dataset,
+ ) = return_subsets(self.dataset)
+
+
+datamodule = DefaultModule