diff --git a/.pre-commit-config.yaml b/.pre-commit-config.yaml
index b7818c75c3a3c31738fb290240ac73bd43d4c280..0fc301b3a02156e97eca69a793215e1527ec8d72 100644
--- a/.pre-commit-config.yaml
+++ b/.pre-commit-config.yaml
@@ -49,11 +49,8 @@ repos:
- id: check-added-large-files
exclude: |
(?x)(
- ^src/ptbench/data/nih_cxr14_re/idiap.json|
- ^src/ptbench/data/nih_cxr14_re/default.json|
- ^src/ptbench/data/padchest/idiap.json|
- ^src/ptbench/data/padchest/no_tb_idiap.json|
- ^src/ptbench/data/padchest/no_tb_idiap.json|
+ ^src/ptbench/data/padchest/idiap.json.bz2|
+ ^src/ptbench/data/padchest/no-tb-idiap.json.bz2|
^tests/data/16bits.png|
^doc/results/img/rad_sign_drop.png
)
diff --git a/pyproject.toml b/pyproject.toml
index ff03e3665b67a566616c10387be8965cf3356ee6..23a5908cf088cb6bf5add2d1d0600c9e7a66fc15 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -220,19 +220,19 @@ hivtb_f9 = "ptbench.data.hivtb.fold_9"
# montgomery-shenzhen-indian-padchest aggregated dataset
mc_ch_in_pc = "ptbench.data.mc_ch_in_pc.default"
-# NIH CXR14 (relabeled)
+# NIH CXR14 (relabeled), multi-class (14 labels)
nih-cxr14 = "ptbench.data.nih_cxr14.default"
nih-cxr14-cardiomegaly = "ptbench.data.nih_cxr14.cardiomegaly"
+# PadChest, multi-class (varied number of labels)
+padchest-idiap = "ptbench.data.padchest.idiap"
+padchest-tb-idiap = "ptbench.data.padchest.tb_idiap"
+padchest-no-tb-idiap = "ptbench.data.padchest.no_tb_idiap"
+padchest-cardiomegaly-idiap = "ptbench.data.padchest.cardiomegaly_idiap"
+
# NIH CXR14 / PadChest aggregated dataset
nih_cxr14_pc_idiap = "ptbench.data.nih_cxr14_re_pc.idiap"
-# PadChest
-padchest_idiap = "ptbench.data.padchest.idiap"
-padchest_tb_idiap = "ptbench.data.padchest.tb_idiap"
-padchest_no_tb_idiap = "ptbench.data.padchest.no_tb_idiap"
-padchest_cm_idiap = "ptbench.data.padchest.cardiomegaly_idiap"
-
[tool.setuptools]
zip-safe = true
package-dir = { "" = "src" }
diff --git a/src/ptbench/data/datamodule.py b/src/ptbench/data/datamodule.py
index 9b5eab61b988e65d5b7d199d1cfcfc854cd784be..0785e2332d61978373dea1a9b536559848a5c117 100644
--- a/src/ptbench/data/datamodule.py
+++ b/src/ptbench/data/datamodule.py
@@ -541,6 +541,11 @@ class ConcatDataModule(lightning.LightningDataModule):
self.set_chunk_size(batch_size, batch_chunk_count)
self.splits = splits
+
+ for dataset_name, split_loaders in splits.items():
+ count = sum([len(k) for k, _ in split_loaders])
+ logger.info(f"Dataset `{dataset_name}` contains {count} samples")
+
self.cache_samples = cache_samples
self._train_sampler = None
self.balance_sampler_by_class = balance_sampler_by_class
diff --git a/src/ptbench/data/image_utils.py b/src/ptbench/data/image_utils.py
index b9f8c8108dd0db18a0cd796b968d26f82da49906..99df51c8ac6b6ac0425c0dbb6013ebf36bc3ab7c 100644
--- a/src/ptbench/data/image_utils.py
+++ b/src/ptbench/data/image_utils.py
@@ -9,26 +9,6 @@ import numpy
import PIL.Image
-class SingleAutoLevel16to8:
- """Converts a 16-bit image to 8-bit representation using "auto-level".
-
- This transform assumes that the input image is gray-scaled.
-
- To auto-level, we calculate the maximum and the minimum of the
- image, and consider such a range should be mapped to the [0,255]
- range of the destination image.
- """
-
- def __call__(self, img):
- imin, imax = img.getextrema()
- irange = imax - imin
- return PIL.Image.fromarray(
- numpy.round(
- 255.0 * (numpy.array(img).astype(float) - imin) / irange
- ).astype("uint8"),
- ).convert("L")
-
-
def remove_black_borders(
img: PIL.Image.Image, threshold: int = 0
) -> PIL.Image.Image:
@@ -47,19 +27,23 @@ def remove_black_borders(
A PIL image with black borders removed
"""
- img = numpy.asarray(img)
+ img_array = numpy.asarray(img)
- if len(img.shape) == 2: # single channel
- mask = numpy.asarray(img) > threshold
- return PIL.Image.fromarray(img[numpy.ix_(mask.any(1), mask.any(0))])
+ if len(img_array.shape) == 2: # single channel
+ mask = numpy.asarray(img_array) > threshold
+ return PIL.Image.fromarray(
+ img_array[numpy.ix_(mask.any(1), mask.any(0))]
+ )
- elif len(img.shape) == 3 and img.shape[2] == 3:
- r_mask = img[:, :, 0] > threshold
- g_mask = img[:, :, 1] > threshold
- b_mask = img[:, :, 2] > threshold
+ elif len(img_array.shape) == 3 and img_array.shape[2] == 3:
+ r_mask = img_array[:, :, 0] > threshold
+ g_mask = img_array[:, :, 1] > threshold
+ b_mask = img_array[:, :, 2] > threshold
mask = r_mask | g_mask | b_mask
- return PIL.Image.fromarray(img[numpy.ix_(mask.any(1), mask.any(0))])
+ return PIL.Image.fromarray(
+ img_array[numpy.ix_(mask.any(1), mask.any(0))]
+ )
else:
raise NotImplementedError
diff --git a/src/ptbench/data/nih_cxr14/__init__.py b/src/ptbench/data/nih_cxr14/__init__.py
index b9954cf126eae1670c87296ad86f0ca6f4f9e758..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 100644
--- a/src/ptbench/data/nih_cxr14/__init__.py
+++ b/src/ptbench/data/nih_cxr14/__init__.py
@@ -1,97 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-"""NIH CXR14 (relabeled) dataset for computer-aided diagnosis.
-
-This dataset was extracted from the clinical PACS database at the National
-Institutes of Health Clinical Center (USA) and represents 60% of all
-their radiographs. It contains labels for fourteen common radiological
-signs in this order: cardiomegaly, emphysema, effusion, hernia, infiltration,
-mass, nodule, atelectasis, pneumothorax, pleural thickening, pneumonia,
-fibrosis, edema and consolidation.
-This is the relabeled version created in the CheXNeXt study.
-
-* Reference: [NIH-CXR14-2017]_
-* Original resolution (height x width or width x height): 1024 x 1024
-* Labels: [CHEXNEXT-2018]_
-* Split reference: [CHEXNEXT-2018]_
-* Protocol ``default``:
-
- * Training samples: 98'637 (including labels)
- * Validation samples: 6'350 (including labels)
- * Test samples: 0
-
-* Protocol `ìdiap``:
- * Images path adapted to Idiap infrastructure
-"""
-
-import importlib.resources
-import os
-
-from ...utils.rc import load_rc
-from ..dataset import JSONDataset
-from ..loader import load_pil_rgb, make_delayed
-
-_protocols = [
- importlib.resources.files(__name__).joinpath("default.json.bz2"),
- importlib.resources.files(__name__).joinpath("cardiomegaly.json.bz2"),
-]
-
-_datadir = load_rc().get("datadir.nih_cxr14_re", os.path.realpath(os.curdir))
-_idiap_folders = load_rc().get("nih_cxr14_re.idiap_folder_structure", False)
-
-
-def _raw_data_loader(sample):
- if _idiap_folders:
- # for folder lookup efficiency, data is split into subfolders
- # each original file is on the subfolder `f[:5]/f`, where f
- # is the original file basename
- basename = os.path.basename(sample["data"])
- return dict(
- data=load_pil_rgb(
- os.path.join(
- _datadir,
- os.path.dirname(sample["data"]),
- basename[:5],
- basename,
- )
- ),
- label=sample["label"],
- )
- else:
- return dict(
- data=load_pil_rgb(os.path.join(_datadir, sample["data"])),
- label=sample["label"],
- )
-
-
-def _loader(context, sample):
- # "context" is ignored in this case - database is homogeneous
- # we returned delayed samples to avoid loading all images at once
- return make_delayed(sample, _raw_data_loader)
-
-
-json_dataset = JSONDataset(
- protocols=_protocols,
- fieldnames=("data", "label"),
- loader=_loader,
-)
-"""NIH CXR14 (relabeled) dataset object."""
-
-
-def _maker(protocol, size=512):
- import torchvision.transforms as transforms
-
- from .. import make_dataset
-
- # ImageNet normalization
- normalize = transforms.Normalize(
- mean=[0.485, 0.456, 0.406], std=[0.229, 0.224, 0.225]
- )
-
- return make_dataset(
- [json_dataset.subsets(protocol)],
- [transforms.Resize((size, size))],
- [transforms.RandomHorizontalFlip()],
- [transforms.ToTensor(), normalize],
- )
diff --git a/src/ptbench/data/padchest/__init__.py b/src/ptbench/data/padchest/__init__.py
index 1151b55f28eb1a6e90310e0abc5c544991b314bb..e69de29bb2d1d6434b8b29ae775ad8c2e48c5391 100644
--- a/src/ptbench/data/padchest/__init__.py
+++ b/src/ptbench/data/padchest/__init__.py
@@ -1,281 +0,0 @@
-# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
-#
-# SPDX-License-Identifier: GPL-3.0-or-later
-"""Padchest dataset for computer-aided diagnosis.
-
-A large chest x-ray image dataset with multi-label annotated reports.
-This dataset includes more than 160,000 images from 67,000 patients that were
-interpreted and reported by radiologists at Hospital San Juan (Spain) from 2009
-to 2017, covering six different position views and additional information on
-image acquisition and patient demography.
-
-We keep only "PA" images here.
-
-* Reference: [PADCHEST-2019]_
-* Original resolution: variable, original size
-* Labels: [PADCHEST-2019]_
-* Split reference: no split
-* Protocol ``default``:
-
- * Training samples: 96'269
- * Validation samples: 0
- * Test samples: 0
-
-* Protocol `ìdiap``:
- * Images path adapted to Idiap infrastructure
-
-* Labels:
-
-['COPD signs',
-'Chilaiditi sign',
-'NSG tube',
-'abnormal foreign body',
-'abscess',
-'adenopathy',
-'air bronchogram',
-'air fluid level',
-'air trapping',
-'alveolar pattern',
-'aortic aneurysm',
-'aortic atheromatosis',
-'aortic button enlargement',
-'aortic elongation',
-'aortic endoprosthesis',
-'apical pleural thickening',
-'artificial aortic heart valve',
-'artificial heart valve',
-'artificial mitral heart valve',
-'asbestosis signs',
-'ascendent aortic elongation',
-'atelectasis',
-'atelectasis basal',
-'atypical pneumonia',
-'axial hyperostosis',
-'azygoesophageal recess shift',
-'azygos lobe',
-'blastic bone lesion',
-'bone cement',
-'bone metastasis',
-'breast mass',
-'bronchiectasis',
-'bronchovascular markings',
-'bullas',
-'calcified adenopathy',
-'calcified densities',
-'calcified fibroadenoma',
-'calcified granuloma',
-'calcified mediastinal adenopathy',
-'calcified pleural plaques',
-'calcified pleural thickening',
-'callus rib fracture',
-'cardiomegaly',
-'catheter',
-'cavitation',
-'central vascular redistribution',
-'central venous catheter',
-'central venous catheter via jugular vein',
-'central venous catheter via subclavian vein',
-'central venous catheter via umbilical vein',
-'cervical rib',
-'chest drain tube',
-'chronic changes',
-'clavicle fracture',
-'consolidation',
-'costochondral junction hypertrophy',
-'costophrenic angle blunting',
-'cyst',
-'dai',
-'descendent aortic elongation',
-'dextrocardia',
-'diaphragmatic eventration',
-'double J stent',
-'dual chamber device',
-'electrical device',
-'emphysema',
-'empyema',
-'end on vessel',
-'endoprosthesis',
-'endotracheal tube',
-'esophagic dilatation',
-'exclude',
-'external foreign body',
-'fibrotic band',
-'fissure thickening',
-'flattened diaphragm',
-'fracture',
-'gastrostomy tube',
-'goiter',
-'granuloma',
-'ground glass pattern',
-'gynecomastia',
-'heart insufficiency',
-'heart valve calcified',
-'hemidiaphragm elevation',
-'hiatal hernia',
-'hilar congestion',
-'hilar enlargement',
-'humeral fracture',
-'humeral prosthesis',
-'hydropneumothorax',
-'hyperinflated lung',
-'hypoexpansion',
-'hypoexpansion basal',
-'increased density',
-'infiltrates',
-'interstitial pattern',
-'kerley lines',
-'kyphosis',
-'laminar atelectasis',
-'lepidic adenocarcinoma',
-'lipomatosis',
-'lobar atelectasis',
-'loculated fissural effusion',
-'loculated pleural effusion',
-'lung metastasis',
-'lung vascular paucity',
-'lymphangitis carcinomatosa',
-'lytic bone lesion',
-'major fissure thickening',
-'mammary prosthesis',
-'mass',
-'mastectomy',
-'mediastinal enlargement',
-'mediastinal mass',
-'mediastinal shift',
-'mediastinic lipomatosis',
-'metal',
-'miliary opacities',
-'minor fissure thickening',
-'multiple nodules',
-'nephrostomy tube',
-'nipple shadow',
-'nodule',
-'non axial articular degenerative changes',
-'normal',
-'obesity',
-'osteopenia',
-'osteoporosis',
-'osteosynthesis material',
-'pacemaker',
-'pectum carinatum',
-'pectum excavatum',
-'pericardial effusion',
-'pleural effusion',
-'pleural mass',
-'pleural plaques',
-'pleural thickening',
-'pneumomediastinum',
-'pneumonia',
-'pneumoperitoneo',
-'pneumothorax',
-'post radiotherapy changes',
-'prosthesis',
-'pseudonodule',
-'pulmonary artery enlargement',
-'pulmonary artery hypertension',
-'pulmonary edema',
-'pulmonary fibrosis',
-'pulmonary hypertension',
-'pulmonary mass',
-'pulmonary venous hypertension',
-'reservoir central venous catheter',
-'respiratory distress',
-'reticular interstitial pattern',
-'reticulonodular interstitial pattern',
-'rib fracture',
-'right sided aortic arch',
-'round atelectasis',
-'sclerotic bone lesion',
-'scoliosis',
-'segmental atelectasis',
-'single chamber device',
-'soft tissue mass',
-'sternoclavicular junction hypertrophy',
-'sternotomy',
-'subacromial space narrowing',
-'subcutaneous emphysema',
-'suboptimal study',
-'superior mediastinal enlargement',
-'supra aortic elongation',
-'surgery',
-'surgery breast',
-'surgery heart',
-'surgery humeral',
-'surgery lung',
-'surgery neck',
-'suture material',
-'thoracic cage deformation',
-'total atelectasis',
-'tracheal shift',
-'tracheostomy tube',
-'tuberculosis',
-'tuberculosis sequelae',
-'unchanged',
-'vascular hilar enlargement',
-'vascular redistribution',
-'ventriculoperitoneal drain tube',
-'vertebral anterior compression',
-'vertebral compression',
-'vertebral degenerative changes',
-'vertebral fracture',
-'volume loss']
-"""
-
-import importlib.resources
-import os
-
-from ...utils.rc import load_rc
-from .. import make_dataset
-from ..dataset import JSONDataset
-from ..loader import load_pil, make_delayed
-
-_protocols = [
- importlib.resources.files(__name__).joinpath("idiap.json.bz2"),
- importlib.resources.files(__name__).joinpath("tb_idiap.json.bz2"),
- importlib.resources.files(__name__).joinpath("no_tb_idiap.json.bz2"),
- importlib.resources.files(__name__).joinpath("cardiomegaly_idiap.json.bz2"),
-]
-
-_datadir = load_rc().get("datadir.padchest", os.path.realpath(os.curdir))
-
-
-def _raw_data_loader(sample):
- return dict(
- data=load_pil(os.path.join(_datadir, sample["data"])),
- label=sample["label"],
- )
-
-
-def _loader(context, sample):
- # "context" is ignored in this case - database is homogeneous
- # we returned delayed samples to avoid loading all images at once
- return make_delayed(sample, _raw_data_loader)
-
-
-json_dataset = JSONDataset(
- protocols=_protocols,
- fieldnames=("data", "label"),
- loader=_loader,
-)
-"""Padchest dataset object."""
-
-
-def _maker(protocol, resize_size=512, cc_size=512, RGB=True):
- import torchvision.transforms as transforms
-
- from ..loader import SingleAutoLevel16to8
-
- post_transforms = []
- if not RGB:
- post_transforms = [transforms.Lambda(lambda x: x.convert("L"))]
-
- return make_dataset(
- [json_dataset.subsets(protocol)],
- [
- SingleAutoLevel16to8(),
- transforms.Resize(resize_size),
- transforms.CenterCrop(cc_size),
- ],
- [transforms.RandomHorizontalFlip()],
- post_transforms,
- )
diff --git a/src/ptbench/data/padchest/cardiomegaly_idiap.json b/src/ptbench/data/padchest/cardiomegaly_idiap.json
new file mode 100644
index 0000000000000000000000000000000000000000..9487ba838f48a22c4f5c1b3f9fd280acbababdf0
--- /dev/null
+++ b/src/ptbench/data/padchest/cardiomegaly_idiap.json
@@ -0,0 +1,164 @@
+{
+ "train": [
+ [
+ "0/128060267991587422580475423001969971705_pjiogk.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/62251368659414281768550499840080640229_8jw2hb.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/30661727075761817007267292459310975718_86nsuj.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/84545123491560549832532130982226930123_vvayv1.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/284353054075043225622260270287627142906_kduagg.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/172366760559200399420370251768770096562_31ewgy.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/87180314458623865269977662447356279507_tcw56c.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/229647061532869405127486256298724729451_o2d0lv.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/110057948387370363009458390264986734403-3_pbni4p.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/110057948387370363009458390264986734403_myst7q.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/13028631954019256091081365260375513084_9b0d5l.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/304673909503207424801802419702123736315_vyckk1.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/163750344543947776171909966951769207897_dgf4vk.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/159216970546428043543933950339353602394_t0rgza.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/220991425902150564296792633654689602163_hrvmt4.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/227808586259025315397189404192472871023_2_ba4iu6.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/289108620230468119174968226927683707287_9wupg2.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/159364525876157332485573893353941089455_aepmh7.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/262822530217543635254713021579502738735_g2v51e.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/172167226340244841253091101374217956877_q2xlp3.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/1256842362861431725328351539259305635_u1qifz.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/126455653911678893263056441159987922516_uz7fwy.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/10996416492353037588312781035930080694_8rstz0.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/76706953811930577985930635244528105325_43rk2l.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/232572031807344965043258496995347066691_g4zxoq.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/184768585622747149847312236491996945269_hpn97q.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/333078556932425505903975417870495109406_19z1qe.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/278147989796380509519073707235611243961_fplpb6.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/278147989796380509519073707235611243961_lgo9hx.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/119688533980562249677219589534851011987_60uvre.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/125061228802875769555301109909833678388_t5xvaz.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/106111337169988693038323274011777746837_0yin7u.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/159127361586483217046684065044208677781_lwjgop.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/267897293741510635554583282472837799427_2_bbiu7z.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/63510470621460583865307590457045621750_wfd1y4.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/199932188422938481238592789979586445399_5tsy8k.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/62774794894109549387630855543283311955_2_tdirok.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/62774794894109549387630855543283311955_9jbiyo.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/245895219126534788779109786698958326807_1a9ait.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ],
+ [
+ "0/250229194886382183353648559275366656935_x3otkq.png",
+ [1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0]
+ ]
+ ]
+}
diff --git a/src/ptbench/data/padchest/cardiomegaly_idiap.json.bz2 b/src/ptbench/data/padchest/cardiomegaly_idiap.json.bz2
deleted file mode 100644
index 52c89dae856d7759b1955d8ece2621aacf5f72c9..0000000000000000000000000000000000000000
Binary files a/src/ptbench/data/padchest/cardiomegaly_idiap.json.bz2 and /dev/null differ
diff --git a/src/ptbench/data/padchest/cardiomegaly_idiap.py b/src/ptbench/data/padchest/cardiomegaly_idiap.py
index d25387388d1540e706876eef2e15543b293113e0..05286fd598569c53d625f0e2513384dac78cdfd4 100644
--- a/src/ptbench/data/padchest/cardiomegaly_idiap.py
+++ b/src/ptbench/data/padchest/cardiomegaly_idiap.py
@@ -1,48 +1,14 @@
# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
#
# SPDX-License-Identifier: GPL-3.0-or-later
-"""Padchest cardiomegaly (idiap protocol) dataset for computer-aided diagnosis.
-
-The first 40 images with cardiomegaly.
-parameters: Label = "Normal", MethodLabel = "Physician", Projection = "PA"
-
-* Split reference: first 100% of cardiomegaly for "train"
-* See :py:mod:`ptbench.data.padchest` for dataset details
-* This configuration resolution: 512 x 512 (default)
-"""
-
-from clapper.logging import setup
-from .. import return_subsets
-from ..base_datamodule import BaseDataModule
-from . import _maker
+from .datamodule import 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,
- ):
- super().__init__(
- train_batch_size=train_batch_size,
- predict_batch_size=predict_batch_size,
- drop_incomplete_batch=drop_incomplete_batch,
- multiproc_kwargs=multiproc_kwargs,
- )
-
- def setup(self, stage: str):
- self.dataset = _maker("cardiomegaly_idiap", RGB=False)
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
+datamodule = DataModule("cardiomegaly_idiap.json")
+"""Padchest cardiomegaly (idiap protocol) dataset for computer-aided diagnosis.
+This split contains the first 40 images with cardiomegaly, with parameters:
+Label = "Normal", MethodLabel = "Physician", Projection = "PA"
-datamodule = DefaultModule
+Read documentation of :py:class:`DataModule` for technical details.
+"""
diff --git a/src/ptbench/data/padchest/datamodule.py b/src/ptbench/data/padchest/datamodule.py
new file mode 100644
index 0000000000000000000000000000000000000000..5db738887196b60ea953c107f7de2603e7a24f02
--- /dev/null
+++ b/src/ptbench/data/padchest/datamodule.py
@@ -0,0 +1,333 @@
+# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
+#
+# SPDX-License-Identifier: GPL-3.0-or-later
+
+import importlib.resources
+import os
+
+import numpy
+import PIL.Image
+
+from torchvision.transforms.functional import center_crop, to_tensor
+
+from ...utils.rc import load_rc
+from ..datamodule import CachingDataModule
+from ..image_utils import remove_black_borders
+from ..split import JSONDatabaseSplit
+from ..typing import DatabaseSplit
+from ..typing import RawDataLoader as _BaseRawDataLoader
+from ..typing import Sample
+
+
+class RawDataLoader(_BaseRawDataLoader):
+ """A specialized raw-data-loader for the PadChest dataset.
+
+ Attributes
+ ----------
+
+ datadir
+ This variable contains the base directory where the database raw data
+ is stored.
+ """
+
+ datadir: str
+
+ def __init__(self):
+ rc = load_rc()
+ self.datadir = rc.get("datadir.padchest", os.path.realpath(os.curdir))
+
+ def sample(self, sample: tuple[str, list[int]]) -> Sample:
+ """Loads a single image sample from the disk.
+
+ Parameters
+ ----------
+
+ sample:
+ A tuple containing the path suffix, within the dataset root folder,
+ where to find the image to be loaded, and an integer, representing the
+ sample label.
+
+
+ Returns
+ -------
+
+ sample
+ The sample representation
+ """
+ # N.B.: PadChest images are encoded as 16-bit grayscale images
+ image = PIL.Image.open(os.path.join(self.datadir, sample[0]))
+ image = remove_black_borders(image)
+ array = numpy.array(image).astype(numpy.float32) / 65536
+ tensor = to_tensor(array)
+ tensor = center_crop(tensor, min(*tensor.shape[1:]))
+
+ # use the code below to view generated images
+ # from torchvision.transforms.functional import to_pil_image
+ # to_pil_image(tensor).show()
+ # __import__("pdb").set_trace()
+
+ return tensor, dict(label=sample[1], name=sample[0]) # type: ignore[arg-type]
+
+ def label(self, sample: tuple[str, list[int]]) -> list[int]:
+ """Loads a single image sample label from the disk.
+
+ Parameters
+ ----------
+
+ sample:
+ A tuple containing the path suffix, within the dataset root folder,
+ where to find the image to be loaded, and an integer, representing the
+ sample label.
+
+
+ Returns
+ -------
+
+ labels
+ The integer labels associated with the sample
+ """
+ return sample[1]
+
+
+def make_split(basename: str) -> DatabaseSplit:
+ """Returns a database split for the NIH CXR-14 database."""
+
+ return JSONDatabaseSplit(
+ importlib.resources.files(__name__.rsplit(".", 1)[0]).joinpath(basename)
+ )
+
+
+class DataModule(CachingDataModule):
+ """Padchest dataset for computer-aided diagnosis.
+
+ A large chest x-ray image dataset with multi-label annotated reports. This
+ dataset includes more than 160,000 images from 67,000 patients that were
+ interpreted and reported by radiologists at Hospital San Juan (Spain) from
+ 2009 to 2017, covering six different position views and additional
+ information on image acquisition and patient demography.
+
+ We keep only postero-anterior "PA" images in our setup.
+
+ * Reference: [PADCHEST-2019]_
+ * Raw data input (on disk):
+
+ * PNG grayscale 16-bit depth images
+ * Resolution: varying resolution
+
+ * Labels: [PADCHEST-2019]_
+ * Output image:
+
+ * Transforms:
+
+ * Load raw 16-bit PNG with :py:mod:`PIL`
+ * Remove excess black borders
+ * Convert image to 32-bit floats between 0. and 1.
+ * Convert to tensor
+ * Center crop, preserving the minimum dimension (height or width)
+
+ * Final specifications
+
+ * Grayscale, encoded as a 1-plane 32-bit float image, square with
+ varying resolutions depending on the raw input image
+ * Labels, in order (some of which may not be present in all splits):
+
+ * COPD signs
+ * Chilaiditi sign
+ * NSG tube
+ * abnormal foreign body
+ * abscess
+ * adenopathy
+ * air bronchogram
+ * air fluid level
+ * air trapping
+ * alveolar pattern
+ * aortic aneurysm
+ * aortic atheromatosis
+ * aortic button enlargement
+ * aortic elongation
+ * aortic endoprosthesis
+ * apical pleural thickening
+ * artificial aortic heart valve
+ * artificial heart valve
+ * artificial mitral heart valve
+ * asbestosis signs
+ * ascendent aortic elongation
+ * atelectasis
+ * atelectasis basal
+ * atypical pneumonia
+ * axial hyperostosis
+ * azygoesophageal recess shift
+ * azygos lobe
+ * blastic bone lesion
+ * bone cement
+ * bone metastasis
+ * breast mass
+ * bronchiectasis
+ * bronchovascular markings
+ * bullas
+ * calcified adenopathy
+ * calcified densities
+ * calcified fibroadenoma
+ * calcified granuloma
+ * calcified mediastinal adenopathy
+ * calcified pleural plaques
+ * calcified pleural thickening
+ * callus rib fracture
+ * cardiomegaly
+ * catheter
+ * cavitation
+ * central vascular redistribution
+ * central venous catheter
+ * central venous catheter via jugular vein
+ * central venous catheter via subclavian vein
+ * central venous catheter via umbilical vein
+ * cervical rib
+ * chest drain tube
+ * chronic changes
+ * clavicle fracture
+ * consolidation
+ * costochondral junction hypertrophy
+ * costophrenic angle blunting
+ * cyst
+ * dai
+ * descendent aortic elongation
+ * dextrocardia
+ * diaphragmatic eventration
+ * double J stent
+ * dual chamber device
+ * electrical device
+ * emphysema
+ * empyema
+ * end on vessel
+ * endoprosthesis
+ * endotracheal tube
+ * esophagic dilatation
+ * exclude
+ * external foreign body
+ * fibrotic band
+ * fissure thickening
+ * flattened diaphragm
+ * fracture
+ * gastrostomy tube
+ * goiter
+ * granuloma
+ * ground glass pattern
+ * gynecomastia
+ * heart insufficiency
+ * heart valve calcified
+ * hemidiaphragm elevation
+ * hiatal hernia
+ * hilar congestion
+ * hilar enlargement
+ * humeral fracture
+ * humeral prosthesis
+ * hydropneumothorax
+ * hyperinflated lung
+ * hypoexpansion
+ * hypoexpansion basal
+ * increased density
+ * infiltrates
+ * interstitial pattern
+ * kerley lines
+ * kyphosis
+ * laminar atelectasis
+ * lepidic adenocarcinoma
+ * lipomatosis
+ * lobar atelectasis
+ * loculated fissural effusion
+ * loculated pleural effusion
+ * lung metastasis
+ * lung vascular paucity
+ * lymphangitis carcinomatosa
+ * lytic bone lesion
+ * major fissure thickening
+ * mammary prosthesis
+ * mass
+ * mastectomy
+ * mediastinal enlargement
+ * mediastinal mass
+ * mediastinal shift
+ * mediastinic lipomatosis
+ * metal
+ * miliary opacities
+ * minor fissure thickening
+ * multiple nodules
+ * nephrostomy tube
+ * nipple shadow
+ * nodule
+ * non axial articular degenerative changes
+ * normal
+ * obesity
+ * osteopenia
+ * osteoporosis
+ * osteosynthesis material
+ * pacemaker
+ * pectum carinatum
+ * pectum excavatum
+ * pericardial effusion
+ * pleural effusion
+ * pleural mass
+ * pleural plaques
+ * pleural thickening
+ * pneumomediastinum
+ * pneumonia
+ * pneumoperitoneo
+ * pneumothorax
+ * post radiotherapy changes
+ * prosthesis
+ * pseudonodule
+ * pulmonary artery enlargement
+ * pulmonary artery hypertension
+ * pulmonary edema
+ * pulmonary fibrosis
+ * pulmonary hypertension
+ * pulmonary mass
+ * pulmonary venous hypertension
+ * reservoir central venous catheter
+ * respiratory distress
+ * reticular interstitial pattern
+ * reticulonodular interstitial pattern
+ * rib fracture
+ * right sided aortic arch
+ * round atelectasis
+ * sclerotic bone lesion
+ * scoliosis
+ * segmental atelectasis
+ * single chamber device
+ * soft tissue mass
+ * sternoclavicular junction hypertrophy
+ * sternotomy
+ * subacromial space narrowing
+ * subcutaneous emphysema
+ * suboptimal study
+ * superior mediastinal enlargement
+ * supra aortic elongation
+ * surgery
+ * surgery breast
+ * surgery heart
+ * surgery humeral
+ * surgery lung
+ * surgery neck
+ * suture material
+ * thoracic cage deformation
+ * total atelectasis
+ * tracheal shift
+ * tracheostomy tube
+ * tuberculosis
+ * tuberculosis sequelae
+ * unchanged
+ * vascular hilar enlargement
+ * vascular redistribution
+ * ventriculoperitoneal drain tube
+ * vertebral anterior compression
+ * vertebral compression
+ * vertebral degenerative changes
+ * vertebral fracture
+ * volume loss
+ """
+
+ def __init__(self, split_filename: str):
+ super().__init__(
+ database_split=make_split(split_filename),
+ raw_data_loader=RawDataLoader(),
+ )
diff --git a/src/ptbench/data/padchest/idiap.json.bz2 b/src/ptbench/data/padchest/idiap.json.bz2
index 8dfa755141a0a2a641e2e1ae2b6cdc972151e3ca..d1594934dcc33507729c3a372e2def7977395975 100644
Binary files a/src/ptbench/data/padchest/idiap.json.bz2 and b/src/ptbench/data/padchest/idiap.json.bz2 differ
diff --git a/src/ptbench/data/padchest/idiap.py b/src/ptbench/data/padchest/idiap.py
index 284a944d0d83bd933123d63732f4f045549084d3..ea8b3dbd24d6fdf770f38c3cde7848687090935c 100644
--- a/src/ptbench/data/padchest/idiap.py
+++ b/src/ptbench/data/padchest/idiap.py
@@ -1,44 +1,17 @@
# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
#
# SPDX-License-Identifier: GPL-3.0-or-later
-"""Padchest (idiap protocol) dataset for computer-aided diagnosis.
-* See :py:mod:`ptbench.data.padchest` for dataset details
-* This configuration resolution: 512 x 512 (default)
-"""
-
-from clapper.logging import setup
-
-from .. import return_subsets
-from ..base_datamodule import BaseDataModule
-from . import _maker
-
-logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
+from .datamodule import DataModule
+datamodule = DataModule("idiap.json.bz2")
+"""Padchest dataset for computer-aided diagnosis (``idiap`` split).
-class DefaultModule(BaseDataModule):
- def __init__(
- self,
- train_batch_size=1,
- predict_batch_size=1,
- drop_incomplete_batch=False,
- multiproc_kwargs=None,
- ):
- super().__init__(
- train_batch_size=train_batch_size,
- predict_batch_size=predict_batch_size,
- drop_incomplete_batch=drop_incomplete_batch,
- multiproc_kwargs=multiproc_kwargs,
- )
+This split contains all images in the database. Read documentation of
+:py:class:`DataModule` for technical details.
- def setup(self, stage: str):
- self.dataset = _maker("idiap")
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
-
-
-datamodule = DefaultModule
+* Split reference: ours
+ * Training samples: 96'269
+ * Validation samples: ?
+ * Test samples: ?
+"""
diff --git a/src/ptbench/data/padchest/no-tb-idiap.json.bz2 b/src/ptbench/data/padchest/no-tb-idiap.json.bz2
new file mode 100644
index 0000000000000000000000000000000000000000..f7b7eefec64a896dc6208dbf887429e1c77e8d99
Binary files /dev/null and b/src/ptbench/data/padchest/no-tb-idiap.json.bz2 differ
diff --git a/src/ptbench/data/padchest/no_tb_idiap.json.bz2 b/src/ptbench/data/padchest/no_tb_idiap.json.bz2
deleted file mode 100644
index 08222cbede330a78a8c6248647b84d4e804d8f97..0000000000000000000000000000000000000000
Binary files a/src/ptbench/data/padchest/no_tb_idiap.json.bz2 and /dev/null differ
diff --git a/src/ptbench/data/padchest/no_tb_idiap.py b/src/ptbench/data/padchest/no_tb_idiap.py
index a424e0fe46890b899804aeac954cb3bd710429ae..62e4b56bee3046c63d13ea7f8cf7bbf848616a96 100644
--- a/src/ptbench/data/padchest/no_tb_idiap.py
+++ b/src/ptbench/data/padchest/no_tb_idiap.py
@@ -1,10 +1,11 @@
# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
#
# SPDX-License-Identifier: GPL-3.0-or-later
-"""Padchest tuberculosis (no TB idiap protocol) dataset for computer-aided
-diagnosis.
-* Protocol ``no TB idiap``:
+from .datamodule import DataModule
+
+datamodule = DataModule("no-tb-idiap.json.bz2")
+"""Padchest dataset for computer-aided diagnosis (``no-tb-idiap`` split).
* Training samples: 20'126
* Validation samples: 1'500
@@ -12,44 +13,21 @@ diagnosis.
* Images path adapted to Idiap infrastructure
-* Labels:
- cardiomegaly, emphysema, effusion, hernia, infiltration,
- mass, nodule, atelectasis, pneumothorax, pleural thickening, pneumonia,
- fibrosis, edema and consolidation
+* Labels (in order):
+
+ * cardiomegaly
+ * emphysema
+ * effusion
+ * hernia
+ * infiltration
+ * mass
+ * nodule
+ * atelectasis
+ * pneumothorax
+ * pleural thickening
+ * pneumonia
+ * fibrosis
+ * edema and consolidation
+
+Read documentation of :py:class:`DataModule` for technical details.
"""
-
-from clapper.logging import setup
-
-from .. import return_subsets
-from ..base_datamodule import BaseDataModule
-from . import _maker
-
-logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
-
-
-class DefaultModule(BaseDataModule):
- def __init__(
- self,
- train_batch_size=1,
- predict_batch_size=1,
- drop_incomplete_batch=False,
- multiproc_kwargs=None,
- ):
- super().__init__(
- train_batch_size=train_batch_size,
- predict_batch_size=predict_batch_size,
- drop_incomplete_batch=drop_incomplete_batch,
- multiproc_kwargs=multiproc_kwargs,
- )
-
- def setup(self, stage: str):
- self.dataset = _maker("no_tb_idiap")
- (
- 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/padchest/tb-idiap.json b/src/ptbench/data/padchest/tb-idiap.json
new file mode 100644
index 0000000000000000000000000000000000000000..cfa02579b34a25d548eef6c184a1e5861ecd93d3
--- /dev/null
+++ b/src/ptbench/data/padchest/tb-idiap.json
@@ -0,0 +1,256 @@
+{
+ "train": [
+ ["0/287829261582195482088446848654649898554_hiv2dj.png", 1],
+ ["0/71781238823267640402215377532202403315_pxkotz.png", 1],
+ ["0/232932494651352886935187910213725327397_542k5d.png", 1],
+ ["0/267611385326465042665315112692206586005_-4xiny.png", 1],
+ ["0/230116782779421561454466256768554783517_cg49ag.png", 1],
+ ["1/189304348002497677549254597676333895442_nr95nw.png", 1],
+ ["1/284083233950421891710386954928396720724_iisntp.png", 1],
+ ["2/62385505932300594213076222597583809293_o4nauv.png", 1],
+ ["3/327929900101827184624358086536821590319_p7mhmy.png", 1],
+ ["3/219525789228894168524584993087468671210_qsb22z.png", 1],
+ ["3/181608309586604103724035528082311350123_hrywbo.png", 1],
+ ["4/198010695012203250912000310773676281863_e8fo5o.png", 1],
+ ["5/94154745525773362638891358070604381275_xu2omi.png", 1],
+ ["5/94154745525773362638891358070604381275_2_mobm9i.png", 1],
+ ["6/149448599615132658944902642295667120353_yhnikj.png", 1],
+ ["7/206757321508676009173922966730015242816_e9ivki.png", 1],
+ ["8/142686622787446619811118950710490862534_t13f05.png", 1],
+ ["8/338359343930739545025011516763914060805_cyusbr.png", 1],
+ ["9/32638629892553023971410131828929042060_0j8zfo.png", 1],
+ ["9/76287465797375088242663014918307149350_7y75c7.png", 1],
+ ["9/30340071673860518284580071098941265573_28fovf.png", 1],
+ ["9/133834049169242825047461122757328574656_hxq15r.png", 1],
+ ["10/130547094371746834245750660340603513752_2czlu0.png", 1],
+ ["10/71032888991710211087565881905995666443_dywshe.png", 1],
+ ["11/196949910603117298426935955787507289138_dnn2xk.png", 1],
+ ["11/333486394263212957144509987040971957354_5ik24x.png", 1],
+ ["11/223149869738399986911371339321970392001_k9oayh.png", 1],
+ ["54/28753966717384878552172419373198893194_z3a3sb.png", 1],
+ ["54/214014914699503892891500996873152128828_wbgkx9.png", 1],
+ ["54/195602194021807720669201170455273870892_2_nx9dfb.png", 1],
+ ["12/216840111366964013829543166512013301133058725_02-087-015.png", 1],
+ ["12/216840111366964013686042548532013221094406052_02-000-159.png", 1],
+ ["13/216840111366964013686042548532013189095559443_02-026-060.png", 1],
+ ["18/216840111366964012809176623042010216102115882_02-122-089.png", 1],
+ ["18/216840111366964013590140476722013038115344442_02-068-006.png", 1],
+ ["18/216840111366964013829543166512013349121121083_02-090-198.png", 1],
+ ["19/216840111366964012809176623042010216124251453_02-121-140.png", 1],
+ ["19/216840111366964013829543166512013337132636807_02-092-110.png", 1],
+ ["19/216840111366964013575787468332013022125651679_02-013-087.png", 1],
+ ["19/216840111366964013686042548532013203113624116_02-031-186.png", 1],
+ ["20/216840111366964013076187734852011167042130729_01-012-195.png", 1],
+ ["21/216840111366964013375835044302012149093442693_01-142-098.png", 1],
+ ["21/216840111366964014008416513202014182103505798_01-162-103.png", 1],
+ ["21/216840111366964013307756408102012074125628561_01-091-158.png", 1],
+ ["22/216840111366964013534861372972012334122542714_01-153-110.png", 1],
+ ["22/216840111366964013534861372972012334174219300_01-153-139.png", 1],
+ ["22/216840111366964013076187734852011179190106015_01-020-037.png", 1],
+ ["23/216840111366964013375835044302012145133553460_01-066-089.png", 1],
+ ["23/216840111366964013336287914842012096133241846_01-107-077.png", 1],
+ ["24/216840111366964013217898866992011333091555050_01-027-188.png", 1],
+ ["24/216840111366964013451228379692012281091656994_01-117-103.png", 1],
+ ["25/216840111366964013307756408102012065122438760_01-086-131.png", 1],
+ ["26/216840111366964013274515230432012025083713232_01-043-028.png", 1],
+ ["27/216840111366964013274515230432012027125506724_01-097-014.png", 1],
+ ["28/216840111366964013962490064942014118114434097_01-188-109.png", 1],
+ ["28/216840111366964013076187734852011172122834037_01-021-101.png", 1],
+ ["29/216840111366964013402131755672012172160839897_01-136-125.png", 1],
+ ["29/216840111366964013534861372972012345121350712_01-136-073.png", 1],
+ ["29/216840111366964013217898866992012016124428082_01-102-080.png", 1],
+ ["31/216840111366964013962490064942014125151556277_01-172-070.png", 1],
+ ["32/216840111366964012558082906712010126095539029_03-150-186.png", 1],
+ ["32/216840111366964012734950068292010132105523907_03-150-117.png", 1],
+ ["32/216840111366964012819207061112010251103013727_03-130-173.png", 1],
+ ["32/216840111366964012768025509942010193092643957_03-140-129.png", 1],
+ ["33/216840111366964012734950068292010144122120771_03-148-189.png", 1],
+ ["33/216840111366964012922382741642010364091957451_03-196-136.png", 1],
+ ["34/216840111366964012558082906712010123085205684_03-146-167.png", 1],
+ ["35/216840111366964012558082906712010096135413767_03-182-119.png", 1],
+ ["36/216840111366964012922382741642010361094534387_03-195-044.png", 1],
+ ["36/216840111366964012283393834152009029134933991_00-010-044.png", 1],
+ ["37/216840111366964012989926673512011131130808585_00-155-122.png", 1],
+ ["37/216840111366964012283393834152009026094739380_00-012-198.png", 1],
+ ["37/216840111366964012373310883942009201083307239_00-029-101.png", 1],
+ ["37/216840111366964012373310883942009201094902325_00-029-010.png", 1],
+ ["37/216840111366964012373310883942009201110112337_00-029-103.png", 1],
+ ["37/216840111366964012373310883942009205080440759_00-026-186.png", 1],
+ ["37/216840111366964012373310883942009187120837458_00-033-022.png", 1],
+ ["37/216840111366964013076187734852011200083643751_00-142-079.png", 1],
+ ["38/216840111366964013076187734852011272200326669_00-134-107.png", 1],
+ ["38/216840111366964013076187734852011196103810560_00-144-029.png", 1],
+ ["38/216840111366964013076187734852011193131842486_00-144-131.png", 1],
+ ["38/216840111366964012989926673512011145084722078_00-150-081.png", 1],
+ ["38/216840111366964012959786098432011042111403479_00-180-113.png", 1],
+ ["38/216840111366964012283393834152009034124021993_00-056-037.png", 1],
+ ["38/216840111366964012558082906712009299150550234_00-080-099.png", 1],
+ ["39/216840111366964012373310883942009146100514519_00-040-128.png", 1],
+ ["39/216840111366964012373310883942009145120705372_00-036-050.png", 1],
+ ["39/216840111366964012487858717522009272091546479_00-020-094.png", 1],
+ ["39/216840111366964012487858717522009219105038840_00-001-025.png", 1],
+ ["39/216840111366964012487858717522009219113307532_00-001-048.png", 1],
+ ["39/216840111366964012373310883942009092152630829_00-069-050.png", 1],
+ ["40/216840111366964012339356563862009047140513629_00-053-012.png", 1],
+ ["40/216840111366964012373310883942009145124513316_00-035-198.png", 1],
+ ["40/1275224347932024878238_00-099-157.png", 1],
+ ["42/216840111366964013076187734852011221130705130_00-114-119.png", 1],
+ ["43/216840111366964012959786098432011040111031806_00-173-078.png", 1],
+ ["43/216840111366964012959786098432011040111031806_00-173-030.png", 1],
+ ["43/216840111366964012959786098432011040111031806_00-173-025.png", 1],
+ ["43/216840111366964012558082906712009323164641947_00-106-182.png", 1],
+ ["43/216840111366964013076187734852011185152425386_00-146-061.png", 1],
+ ["0/20536686640136348236148679891455886468_k6ga29.png", 0],
+ ["0/238285621348398466668514178112618553012_a7k6dv.png", 0],
+ ["0/152191969602076825998375638267191596461_ck9qkz.png", 0],
+ ["0/84289138814897824414670894017743137660_xjsw7k.png", 0],
+ ["0/258055173326723517262722952856353517225_gxjwwm.png", 0],
+ ["0/243837679625088918208093479515843617254_znbkkx.png", 0],
+ ["0/337234085099221143119428528401779812468_1hc03o.png", 0],
+ ["0/12582988208641810053552022080964044412_bdbfde.png", 0],
+ ["0/120005774206062253919592068222866365316_5rzk13.png", 0],
+ ["0/51753302240077846214410306432856713330_w0bmgz.png", 0],
+ ["0/151619064599028082995709172863126236342_adr7uc.png", 0],
+ ["0/97965053291113766610970450200683915158_ta7i4n.png", 0],
+ ["0/182408105675910680785179281207852452925_ybx7hr.png", 0],
+ ["0/319976350840661489788378306853554168892_is6ipx.png", 0],
+ ["0/173348027872651650413607534885352239722_ksfbkb.png", 0],
+ ["0/267593312931260619142226905522973356507_dfimnx.png", 0],
+ ["0/105591880985588572332898967525311229749_yqjtuk.png", 0],
+ ["0/330324745759424821132494701991767497957_9fc9au.png", 0],
+ ["0/175828040998318011818448557383860207677_m5a8ih.png", 0],
+ ["0/101408931302974623103956952273007774313_9stb1t.png", 0],
+ ["0/316808028365561931793158511235924256690_01sv6w.png", 0],
+ ["0/121722089111950575741877258740783632351_55kku6.png", 0],
+ ["0/257363608869672643452639640460584331841_2m9azj.png", 0],
+ ["0/278948104568698034857207837457891721065_z574go.png", 0],
+ ["0/153144722300895942937262007594583779219_ohw52d.png", 0],
+ ["0/125998750830791689348030012609205089696_r0z2uf.png", 0],
+ ["0/49951034854725910980631707796081134817_38f62l.png", 0],
+ ["0/316487933899836980515723069532430938491_jsr618.png", 0],
+ ["0/160591118781379370017611727315537422584_8sb33g.png", 0],
+ ["0/126021751903391108689458989626008942485_26b1rr.png", 0],
+ ["0/49031120128502913776108011441053379453_3qyvw8.png", 0],
+ ["0/319750193216879761537585377422638859654_tq1iik.png", 0],
+ ["0/42728540250863331876520898382409588161_o4q8sm.png", 0],
+ ["0/20450229093471036935089654431118253290_fsak6z.png", 0],
+ ["0/237085130674557205383750650934770251154_kr2cfo.png", 0],
+ ["0/4378551502088567463362725870936911970_tcb27t.png", 0],
+ ["0/235156922651079386633702235134872520572_1kcee2.png", 0],
+ ["0/244098485281969800050011620079615032187_kmiw7o.png", 0],
+ ["0/15257471714388675949620391557999378309_vyin8e.png", 0],
+ ["0/330657416961603932280429915924322632486_umk9rt.png", 0],
+ ["0/223182426517840299226527275743732068184_spng8r.png", 0],
+ ["0/223805166479327036969696654106527616531_2_t9llp8.png", 0],
+ ["0/187748529663266333212518173684958471015_ittz4x.png", 0],
+ ["0/52863917336388719685790911566203013345_8ctt1b.png", 0],
+ ["0/298250519825265668757842559853876298941_d6js3t.png", 0],
+ ["0/53271040340917494534931731766099840496_bzweu7.png", 0],
+ ["0/12130328430935630448520149687013276863_99k39n.png", 0],
+ ["0/62099871704902880343187697364243424587_5pzx3y.png", 0],
+ ["0/65266590064606711020514044534829206992_emiz6y.png", 0],
+ ["0/203867956156632398461581965823374154403_hcfaj6.png", 0],
+ ["0/239027903366999274333251661962927457884_0f0trw.png", 0],
+ ["0/306704549447961580693205362675398392203_lug40p.png", 0],
+ ["0/306704549447961580693205362675398392203_3_qxsdo7.png", 0],
+ ["0/43974812568761155667942095366465446895_3ninvk.png", 0],
+ ["0/74576479405710985759203367628923301316_i9epl9.png", 0],
+ ["0/130110969611250982875259631791477091714_jxnrnp.png", 0],
+ ["0/150762056036635610657089202556643673250_rczqej.png", 0],
+ ["0/255293910021372270231348852078918972406_gczlpx.png", 0],
+ ["0/48039780599266839142703323004089311640_jui3wg.png", 0],
+ ["0/176702947772428028299305495478494543149_wkumlt.png", 0],
+ ["0/339378129561197015481434844854643771030_uq3k6q.png", 0],
+ ["0/287026865546374889085157036246558460511_l0lxwq.png", 0],
+ ["0/172555436098471526561614627462201474466_6rvch6.png", 0],
+ ["0/79355438231207002902084777212886277284_k1cyrv.png", 0],
+ ["0/314277032586409115721922718327642931728_vi0hn8.png", 0],
+ ["0/291629452559210611519089695617689344215_c9p3gg.png", 0],
+ ["0/25528567975108302073238456043150484827_d4odva.png", 0],
+ ["0/61569799595446412488213405581720832033_b84v2f.png", 0],
+ ["0/200862359831761685992552819450233793891_z09t4i.png", 0],
+ ["0/75078999225025048286726158739709244898_clh5ow.png", 0],
+ ["0/159690112140413823244967667566418197322_nrua7y.png", 0],
+ ["0/334589562203518771379665149239316362968_3nh7ym.png", 0],
+ ["0/291366571583131755848449851925170824781_ce0bx6.png", 0],
+ ["0/305453048517584511909927361247180991655_2x71z9.png", 0],
+ ["0/44536207964452859015038957006687811448_noeqf7.png", 0],
+ ["0/214260817085057432303623958907855926368_jrl28m.png", 0],
+ ["0/170868574132883919470570081008800440313_0zq8qv.png", 0],
+ ["0/28041611833871250677444368303361730215_ff13fg.png", 0],
+ ["0/90464809371777154323219683309097631033_g7ks8z.png", 0],
+ ["0/339511729523635430557303404954216047272_rwrvt6.png", 0],
+ ["0/26244571749755521880111325776337650586_zofmec.png", 0],
+ ["0/26244571749755521880111325776337650586_0mezs6.png", 0],
+ ["0/169811598307650468944273653237225758377_ri69fd.png", 0],
+ ["0/72598842051928771977753725274938862160_2pamb0.png", 0],
+ ["0/229495322771160678933819106193415737007_6298bo.png", 0],
+ ["0/265767708505416563059027799932479206802_purml1.png", 0],
+ ["0/255305021945238682284369757239439639240_1gk2z6.png", 0],
+ ["0/208887182360953191530026578700363773829_qgx13z.png", 0],
+ ["0/188512239145790616137268274285581986225_efhnpk.png", 0],
+ ["0/323279587618131957350287102063368542386_269oky.png", 0],
+ ["0/28387517755754605410664016700168547553_ui2lcv.png", 0],
+ ["0/77914334473380412364913757909337796966_x0exbf.png", 0],
+ ["0/35040582196710953341366337559603278309_mks08j.png", 0],
+ ["0/32078055345180912106051534833637957932_fkcmpo.png", 0],
+ ["0/215216312896434609819729304269555722027_9i3jyg.png", 0],
+ ["0/14747842373519304109072114951815645766_kjxsxw.png", 0],
+ ["0/300765837088870057523971311706854661569_55ii9c.png", 0],
+ ["0/102102747557522113902850993567166035648_-5a78l.png", 0],
+ ["0/47481551763215566757392610054682744939_3azhzs.png", 0],
+ ["0/125766228454316130624782540198484575895_9ha77d.png", 0]
+ ],
+ "test": [
+ ["44/216840111366964012373310883942009170111625797_00-097-039.png", 1],
+ ["44/216840111366964012373310883942009202102102178_00-034-039.png", 1],
+ ["44/216840111366964012373310883942009187123652081_00-032-157.png", 1],
+ ["44/216840111366964012373310883942009187171059045_00-032-135.png", 1],
+ ["44/216840111366964012373310883942009187120903052_00-032-141.png", 1],
+ ["44/216840111366964012373310883942009187120813381_00-032-175.png", 1],
+ ["44/216840111366964012373310883942009082135558614_00-049-056.png", 1],
+ ["45/216840111366964012487858717522009218100750234_00-005-054.png", 1],
+ ["45/216840111366964012373310883942009117084022290_00-063-162.png", 1],
+ ["45/216840111366964012487858717522009231091732422_00-003-050.png", 1],
+ ["45/216840111366964012373310883942009159112500535_00-093-077.png", 1],
+ ["45/216840111366964012373310883942009132135802286_00-041-118.png", 1],
+ ["45/216840111366964012558082906712009302095926650_00-078-157.png", 1],
+ ["45/216840111366964012959786098432011038093625941_00-174-028.png", 1],
+ ["45/216840111366964012989926673512011097174442376_00-183-163.png", 1],
+ ["46/216840111366964012283393834152009020133704931_00-014-063.png", 1],
+ ["47/1275224347932024875489_00-095-072.png", 1],
+ ["47/216840111366964012487858717522009281094308471_00-073-123.png", 1],
+ ["47/216840111366964012373310883942009159103757279_00-092-099.png", 1],
+ ["47/216840111366964013076187734852011187170414896_00-145-078.png", 1],
+ ["47/216840111366964013076187734852011187170414896_00-145-192.png", 1],
+ ["47/12752243479320241414701_00-171-104.png", 1],
+ ["47/216840111366964012989926673512011097174442376_00-184-039.png", 1],
+ ["48/216840111366964012922382741642010355113626877_04-000-161.png", 1],
+ ["49/216840111366964012819207061112010321104541236_04-018-166.png", 1],
+ ["0/233506623954258136177168751019496286188_muxzw8.png", 0],
+ ["0/257273696639795055440516119201724869708_szlui2.png", 0],
+ ["0/126887219835336055655888692181690517955_1uyqv0.png", 0],
+ ["0/234216022935307048216299336991419380952_sgrlgf.png", 0],
+ ["0/335952795723413338158135725278292632828_iepbz5.png", 0],
+ ["0/82090431388093551321398783174196866128_tr2g8h.png", 0],
+ ["0/198687427221787561855370297023668306242_260owe.png", 0],
+ ["0/68579354598156854599040192419227509037_koc7xf.png", 0],
+ ["0/144478523611315461003350014835868422251_2_s8dluj.png", 0],
+ ["0/302929243006484152942115282128804026168_n68554.png", 0],
+ ["0/164834710188971445829128771258574408035_121u4x.png", 0],
+ ["0/89135767289800763526891535832874263112_j11bkl.png", 0],
+ ["0/289813269991598332532654906677599208958_u5qqk4.png", 0],
+ ["0/164997689015514390292649262625407524157_5p3hbr.png", 0],
+ ["0/284969877609993327035605642598902430366_71qh8x.png", 0],
+ ["0/183325708478983298834864550015860941024_yo7ozp.png", 0],
+ ["0/168265719116971404186158893567526651554_bu6ryj.png", 0],
+ ["0/143131945457640249494238411094875089491_nb0guj.png", 0],
+ ["0/242348865558077717792572639139276481233_tb3asf.png", 0],
+ ["0/200494933217168186622620771086149301425_xwagt4.png", 0],
+ ["0/217342025624397376056025865389719461061_0s1x0d.png", 0],
+ ["0/339445012265588569253208175476185397432_y7ahfu.png", 0],
+ ["0/91815483074912963411621985384575850871_-ja50t.png", 0],
+ ["0/215041224245905944888939370272745392061_993y3a.png", 0],
+ ["0/175965847113500546623993767251681877654_9dyncn.png", 0]
+ ]
+}
diff --git a/src/ptbench/data/padchest/tb_idiap.json.bz2 b/src/ptbench/data/padchest/tb_idiap.json.bz2
deleted file mode 100644
index 99d0bb971acef3ee860be789209985da6d96fed3..0000000000000000000000000000000000000000
Binary files a/src/ptbench/data/padchest/tb_idiap.json.bz2 and /dev/null differ
diff --git a/src/ptbench/data/padchest/tb_idiap.py b/src/ptbench/data/padchest/tb_idiap.py
index 1a2516202e676386d149fbfc684e3b93e805978e..69c2708b6872c6a2dbc3a71ab88d2d41b9e4c5e1 100644
--- a/src/ptbench/data/padchest/tb_idiap.py
+++ b/src/ptbench/data/padchest/tb_idiap.py
@@ -1,48 +1,15 @@
# SPDX-FileCopyrightText: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
#
# SPDX-License-Identifier: GPL-3.0-or-later
-"""Padchest tuberculosis (idiap protocol) dataset for computer-aided diagnosis.
-The 125 healthy images are the first 125 padchest images with the following
-parameters: Label = "Normal", MethodLabel = "Physician", Projection = "PA"
+from .datamodule import DataModule
-* Split reference: first 80% of TB and healthy CXR for "train", rest for "test"
-* See :py:mod:`ptbench.data.padchest` for dataset details
-* This configuration resolution: 512 x 512 (default)
-"""
-
-from clapper.logging import setup
-
-from .. import return_subsets
-from ..base_datamodule import BaseDataModule
-from . import _maker
-
-logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
+datamodule = DataModule("tb-idiap.json")
+"""Padchest dataset for computer-aided diagnosis (``tb-idiap`` split).
+This split contains 125 healthy images are the first 125 padchest images with
+the following parameters: Label = "Normal", MethodLabel = "Physician",
+Projection = "PA"
-class DefaultModule(BaseDataModule):
- def __init__(
- self,
- train_batch_size=1,
- predict_batch_size=1,
- drop_incomplete_batch=False,
- multiproc_kwargs=None,
- ):
- super().__init__(
- train_batch_size=train_batch_size,
- predict_batch_size=predict_batch_size,
- drop_incomplete_batch=drop_incomplete_batch,
- multiproc_kwargs=multiproc_kwargs,
- )
-
- def setup(self, stage: str):
- self.dataset = _maker("tb_idiap", RGB=False)
- (
- self.train_dataset,
- self.validation_dataset,
- self.extra_validation_datasets,
- self.predict_dataset,
- ) = return_subsets(self.dataset)
-
-
-datamodule = DefaultModule
+Read documentation of :py:class:`DataModule` for technical details.
+"""