Merge branch 'add-datamodule' of gitlab.idiap.ch:biosignal/software/ptbench...

Merge branch 'add-datamodule' of gitlab.idiap.ch:biosignal/software/ptbench into add-datamodule-gradcam

.pre-commit-config.yaml

@@ -7,7 +7,7 @@
 # See https://pre-commit.com/hooks.html for more hooks
 repos:
   - repo: https://github.com/psf/black
-    rev: 23.7.0
+    rev: 23.9.1
     hooks:
       - id: black
   - repo: https://github.com/pycqa/docformatter
@@ -23,7 +23,7 @@ repos:
     hooks:
       - id: flake8
   - repo: https://github.com/pre-commit/mirrors-mypy
-    rev: v1.4.1
+    rev: v1.5.1
     hooks:
     - id: mypy
       args: [
@@ -33,7 +33,7 @@ repos:
         --ignore-missing-imports,
       ]
   - repo: https://github.com/asottile/pyupgrade
-    rev: v3.10.1
+    rev: v3.11.0
     hooks:
     - id: pyupgrade
       args: [--py39-plus]

src/ptbench/config/data/tbx11k/datamodule.py

@@ -87,7 +87,11 @@ class RawDataLoader(_BaseRawDataLoader):
         # to_pil_image(tensor).show()
         # __import__("pdb").set_trace()
 
-        return tensor, dict(label=sample[1], name=sample[0])  # type: ignore[arg-type]
+        return tensor, dict(
+            label=sample[1],
+            name=sample[0],
+            radsign_bboxes=self.bbox_annotations(sample),
+        )
 
     def label(self, sample: DatabaseSample) -> int:
         """Loads a single image sample label from the disk.

src/ptbench/data/datamodule.py

@@ -528,11 +528,11 @@ class ConcatDataModule(lightning.LightningDataModule):
         self.parallel = parallel  # immutable, otherwise would need to call
 
         self.pin_memory = (
-            torch.cuda.is_available() or torch.backends.mps.is_available()
+            torch.cuda.is_available() or torch.backends.mps.is_available()  # type: ignore
         )  # should only be true if GPU available and using it
 
         # datasets that have been setup() for the current stage
-        self._datasets: CachingDataModule.DatasetDictionary = {}
+        self._datasets: ConcatDataModule.DatasetDictionary = {}
 
     @property
     def parallel(self) -> int:

tests/test_tbx11k.py

@@ -7,6 +7,7 @@ import importlib
 import typing
 
 import pytest
+import torch
 
 
 def id_function(val):
@@ -147,6 +148,71 @@ def test_protocol_consistency(
     )
 
 
+def check_loaded_batch(
+    batch,
+    batch_size: int,
+    prefixes: typing.Sequence[str],
+):
+    """Checks the consistence of an individual (loaded) batch.
+
+    Parameters
+    ----------
+
+    batch
+        The loaded batch to be checked.
+
+    size
+        The mini-batch size
+    """
+
+    assert len(batch) == 2  # data, metadata
+
+    assert isinstance(batch[0], torch.Tensor)
+    assert batch[0].shape[0] == batch_size  # mini-batch size
+    assert batch[0].shape[1] == 3  # grayscale images
+    assert batch[0].shape[2] == batch[0].shape[3]  # image is square
+    assert batch[0].shape[2] == 512  # image is 512 pixels large
+
+    assert isinstance(batch[1], dict)  # metadata
+    assert (
+        len(batch[1]) == 3
+    )  # label, name and radiological sign bounding-boxes
+
+    assert "label" in batch[1]
+    assert all([k in (0, 1) for k in batch[1]["label"]])
+
+    assert "name" in batch[1]
+    assert all(
+        [any([k.startswith(j) for j in prefixes]) for k in batch[1]["name"]]
+    )
+
+    assert "radsign_bboxes" in batch[1]
+
+    for sample, label, bboxes in zip(
+        batch[0], batch[1]["label"], batch[1]["radsign_bboxes"]
+    ):
+        # there must be a sign indicated on the image, if active TB is detected
+        if label == 1:
+            assert len(bboxes[0]) != 0
+
+        # eif label == 0:  # not true, may have TBI!
+        #    assert len(bboxes) == 0
+
+        # asserts all bounding boxes are within the raw image width and height
+        for bbox_label, xmin, ymin, width, height in zip(*bboxes):
+            if label == 1:
+                assert bbox_label == 1
+            else:
+                assert bbox_label == 0
+            assert (xmin + width) < sample.shape[2]
+            assert (ymin + height) < sample.shape[1]
+
+    # use the code below to view generated images
+    # from torchvision.transforms.functional import to_pil_image
+    # to_pil_image(batch[0][0]).show()
+    # __import__("pdb").set_trace()
+
+
 @pytest.mark.skip_if_rc_var_not_set("datadir.tbx11k")
 @pytest.mark.parametrize(
     "dataset",
@@ -183,9 +249,7 @@ def test_protocol_consistency(
         ("v2_fold_9", ("imgs/health", "imgs/sick", "imgs/tb")),
     ],
 )
-def test_loading(
-    database_checkers, name: str, dataset: str, prefixes: typing.Sequence[str]
-):
+def test_loading(name: str, dataset: str, prefixes: typing.Sequence[str]):
     datamodule = importlib.import_module(
         f".{name}", "ptbench.config.data.tbx11k"
     ).datamodule
@@ -195,21 +259,13 @@ def test_loading(
 
     loader = datamodule.predict_dataloader()[dataset]
 
-    limit = 3  # limit load checking
+    limit = 50  # limit load checking
     for batch in loader:
         if limit == 0:
             break
-        database_checkers.check_loaded_batch(
+        check_loaded_batch(
             batch,
             batch_size=1,
-            color_planes=3,
             prefixes=prefixes,
-            possible_labels=(0, 1),
         )
         limit -= 1
-
-
-# TODO: Tests for loading bounding boxes:
-# if patient has active tb, then has to have 1 or more bounding boxes
-# if patient does not have active tb, there should be no bounding boxes
-# bounding boxes must be within image (512 x 512 pixels)