.gitlab-ci.yml

@@ -10,6 +10,7 @@ include:
 variables:
   GIT_SUBMODULE_STRATEGY: normal
   GIT_SUBMODULE_DEPTH: 1
+  XDG_CONFIG_HOME: $CI_PROJECT_DIR/tests/data
 
 documentation:
   before_script:

.reuse/dep5

@@ -21,5 +21,6 @@ Files:
  tests/data/*.csv
  tests/data/*.json
  tests/data/*.png
+ tests/*.toml
 Copyright: Copyright © 2023 Idiap Research Institute <contact@idiap.ch>
 License: GPL-3.0-or-later

src/mednet/engine/saliency/viewer.py

@@ -63,7 +63,7 @@ def _overlay_saliency_map(
     Parameters
     ----------
     image
-        The input image that will be overlayed with the saliency map.
+        The input image that will be overlaid with the saliency map.
     saliencies
         The saliency map that will be overlaid on the (raw) image.
     colormap
@@ -118,7 +118,7 @@ def _overlay_bounding_box(
     Parameters
     ----------
     image
-        The input image that will be overlayed with the saliency map.
+        The input image that will be overlaid with the saliency map.
     bbox
         The bounding box to draw on the input image.
     color
@@ -149,13 +149,13 @@ def _process_sample(
     saliencies: numpy.typing.NDArray[numpy.double],
     ground_truth: BoundingBoxes,
 ) -> PIL.Image.Image:
-    """Generate an overlayed representation of the original sample and saliency
+    """Generate an overlaid representation of the original sample and saliency
     maps.
 
     Parameters
     ----------
     raw_data
-        The raw data representing the input sample that will be overlayed with
+        The raw data representing the input sample that will be overlaid with
         saliency maps and annotations.
     saliencies
         The saliency map recovered from the model, that will be imprinted on
@@ -166,7 +166,7 @@ def _process_sample(
     Returns
     -------
     PIL.Image.Image
-        An image with the original raw data overlayed with the different
+        An image with the original raw data overlaid with the different
         elements as selected by the user.
     """
 

src/mednet/models/pasa.py

@@ -75,7 +75,11 @@ class Pasa(pl.LightningModule):
         self.model_transforms = [
             Grayscale(),
             SquareCenterPad(),
-            torchvision.transforms.Resize(512, antialias=True),
+            torchvision.transforms.Resize(
+                512,
+                antialias=True,
+                interpolation=torchvision.transforms.InterpolationMode.BILINEAR,
+            ),
         ]
 
         self._train_loss = train_loss

src/mednet/scripts/experiment.py

@@ -67,7 +67,6 @@ def experiment(
 
     save_sh_command(output_folder / "command.sh")
 
-    # training
     logger.info("Started training")
 
     from .train import train
@@ -139,3 +138,35 @@ def experiment(
     )
 
     logger.info("Ended evaluating")
+
+    logger.info("Started generating saliencies")
+
+    from .saliency.generate import generate
+
+    saliencies_gen_folder = output_folder / "gradcam" / "saliencies"
+
+    ctx.invoke(
+        generate,
+        model=model,
+        datamodule=datamodule,
+        weight=train_output_folder,
+        output_folder=saliencies_gen_folder,
+    )
+
+    logger.info("Ended generating saliencies")
+
+    logger.info("Started viewing saliencies")
+
+    from .saliency.view import view
+
+    saliencies_view_folder = output_folder / "gradcam" / "visualizations"
+
+    ctx.invoke(
+        view,
+        model=model,
+        datamodule=datamodule,
+        input_folder=saliencies_gen_folder,
+        output_folder=saliencies_view_folder,
+    )
+
+    logger.info("Ended viewing saliencies")

src/mednet/scripts/saliency/interpretability.py

@@ -23,10 +23,19 @@ logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
 
    .. code:: sh
 
-      mednet saliency interpretability -vv tbx11k-v1-healthy-vs-atb --input-folder=parent-folder/saliencies/ --output-json=path/to/interpretability-scores.json
+      mednet saliency interpretability -vv pasa tbx11k-v1-healthy-vs-atb --input-folder=parent-folder/saliencies/ --output-json=path/to/interpretability-scores.json
 
 """,
 )
+@click.option(
+    "--model",
+    "-m",
+    help="""A lightning module instance implementing the network architecture
+    (not the weights, necessarily) to be used for inference.  Currently, only
+    supports pasa and densenet models.""",
+    required=True,
+    cls=ResourceOption,
+)
 @click.option(
     "--datamodule",
     "-d",
@@ -78,6 +87,7 @@ logger = setup(__name__.split(".")[0], format="%(levelname)s: %(message)s")
 )
 @verbosity_option(logger=logger, cls=ResourceOption, expose_value=False)
 def interpretability(
+    model,
     datamodule,
     input_folder,
     target_label,
@@ -114,7 +124,7 @@ def interpretability(
 
     from ...engine.saliency.interpretability import run
 
-    datamodule.model_transforms = []
+    datamodule.model_transforms = model.transforms
     datamodule.prepare_data()
     datamodule.setup(stage="predict")
 

tests/conftest.py

@@ -2,16 +2,14 @@
 #
 # SPDX-License-Identifier: GPL-3.0-or-later
 
-import os
 import pathlib
-import tempfile
 import typing
-import zipfile
 
+import numpy
 import pytest
-import tomli_w
 import torch
 
+from mednet.data.split import JSONDatabaseSplit
 from mednet.data.typing import DatabaseSplit
 
 
@@ -97,55 +95,6 @@ def temporary_basedir(tmp_path_factory):
     return tmp_path_factory.mktemp("test-cli")
 
 
-def pytest_sessionstart(session: pytest.Session) -> None:
-    """Preset the session start to ensure the Montgomery dataset is always available.
-
-    Parameters
-    ----------
-    session
-        The session to use.
-    """
-
-    from mednet.utils.rc import load_rc
-
-    rc = load_rc()
-
-    database_dir = rc.get("datadir.montgomery")
-    if database_dir is not None:
-        # if the user downloaded it, use that copy
-        return
-
-    # else, we must extract the LFS component (we are likely on the CI)
-    archive = (
-        pathlib.Path(__file__).parents[0] / "data" / "lfs" / "test-database.zip"
-    )
-    assert archive.exists(), (
-        f"Neither datadir.montgomery is set on the global configuration, "
-        f"(typically ~/.config/mednet.toml), or it is possible to detect "
-        f"the presence of {archive}' (did you git submodule init --update "
-        f"this submodule?)"
-    )
-
-    montgomery_tempdir = tempfile.TemporaryDirectory()
-    rc.setdefault("datadir.montgomery", montgomery_tempdir.name)
-
-    with zipfile.ZipFile(archive) as zf:
-        zf.extractall(montgomery_tempdir.name)
-
-    config_filename = "mednet.toml"
-    with open(
-        os.path.join(montgomery_tempdir.name, config_filename), "wb"
-    ) as f:
-        tomli_w.dump(rc.data, f)
-        f.flush()
-
-    os.environ["XDG_CONFIG_HOME"] = montgomery_tempdir.name
-
-    # stash the newly created temporary directory so we can erase it when the
-    key = pytest.StashKey[tempfile.TemporaryDirectory]()
-    session.stash[key] = montgomery_tempdir
-
-
 class DatabaseCheckers:
     """Helpers for database tests."""
 
@@ -156,13 +105,14 @@ class DatabaseCheckers:
         prefixes: typing.Sequence[str],
         possible_labels: typing.Sequence[int],
     ):
-        """Run a simple consistence check on the data split.
+        """Run a simple consistency check on the data split.
 
         Parameters
         ----------
         split
             An instance of DatabaseSplit.
         lengths
+
             A dictionary that contains keys matching those of the split (this will
             be checked).  The values of the dictionary should correspond to the
             sizes of each of the datasets in the split.
@@ -197,12 +147,13 @@ class DatabaseCheckers:
         color_planes: int,
         prefixes: typing.Sequence[str],
         possible_labels: typing.Sequence[int],
+        expected_num_labels: int,
+        expected_image_shape: typing.Optional[tuple[int, ...]] = None,
     ):
-        """Check the consistence of an individual (loaded) batch.
+        """Check the consistency of an individual (loaded) batch.
 
         Parameters
         ----------
-
         batch
             The loaded batch to be checked.
         batch_size
@@ -214,14 +165,22 @@ class DatabaseCheckers:
             prefixes.
         possible_labels
             These are the list of possible labels contained in any split.
+        expected_num_labels
+            The expected number of labels each sample should have.
+        expected_image_shape
+            The expected shape of the image (num_channels, width, height).
         """
 
         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] == color_planes  # grayscale images
-        assert batch[0].shape[2] == batch[0].shape[3]  # image is square
+        assert batch[0].shape[1] == color_planes
+
+        if expected_image_shape:
+            assert all(
+                [data.shape == expected_image_shape for data in batch[0]]
+            )
 
         assert isinstance(batch[1], dict)  # metadata
         assert len(batch[1]) == 2  # label and name
@@ -229,6 +188,9 @@ class DatabaseCheckers:
         assert "label" in batch[1]
         assert all([k in possible_labels for k in batch[1]["label"]])
 
+        if expected_num_labels:
+            assert len(batch[1]["label"]) == expected_num_labels
+
         assert "name" in batch[1]
         assert all(
             [any([k.startswith(j) for j in prefixes]) for k in batch[1]["name"]]
@@ -239,6 +201,56 @@ class DatabaseCheckers:
         # to_pil_image(batch[0][0]).show()
         # __import__("pdb").set_trace()
 
+    @staticmethod
+    def check_image_quality(
+        datamodule,
+        reference_histogram_file,
+        compare_type="equal",
+        pearson_coeff_threshold=0.005,
+    ):
+        ref_histogram_splits = JSONDatabaseSplit(reference_histogram_file)
+
+        for split_name in ref_histogram_splits:
+            raw_samples = datamodule.splits[split_name][0][0]
+
+            # It is not possible to get a sample from a Dataset by name/path, only by index.
+            # This creates a dict of sample name to dataset index.
+            raw_samples_indices = {}
+            for idx, rs in enumerate(raw_samples):
+                raw_samples_indices[rs[0]] = idx
+
+            for ref_hist_path, ref_hist_data in ref_histogram_splits[
+                split_name
+            ]:
+                # Get index in the dataset that will return the data corresponding to the specified sample name
+                dataset_sample_index = raw_samples_indices[ref_hist_path]
+
+                image_tensor = datamodule._datasets[split_name][
+                    dataset_sample_index
+                ][0]
+
+                histogram = []
+                for color_channel in image_tensor:
+                    color_channel = numpy.multiply(
+                        color_channel.numpy(), 255
+                    ).astype(int)
+                    histogram.extend(
+                        numpy.histogram(
+                            color_channel, bins=256, range=(0, 256)
+                        )[0].tolist()
+                    )
+
+                if compare_type == "statistical":
+                    # Compute pearson coefficients between histogram and reference
+                    # and check the similarity within a certain threshold
+                    pearson_coeffs = numpy.corrcoef(histogram, ref_hist_data)
+                    assert (
+                        1 - pearson_coeff_threshold <= pearson_coeffs[0][1] <= 1
+                    )
+
+                else:
+                    assert histogram == ref_hist_data
+
 
 @pytest.fixture
 def database_checkers():

tests/data/histograms/raw_data/histograms_padchest_idiap.json

+{
+  "train": [
+    ["43/216840111366964012558082906712009341094009402_00-084-033.png", [10737, 236, 243, 283, 266, 289, 465, 564, 527, 609, 596, 713, 675, 602, 698, 668, 654, 581, 619, 770, 717, 868, 798, 877, 914, 752, 975, 1168, 1238, 1515, 1247, 1446, 1658, 1754, 1853, 2385, 2756, 2908, 3404, 4017, 4205, 5651, 6197, 7375, 8072, 8391, 10084, 11275, 11945, 11665, 14699, 15124, 14660, 16467, 15130, 17940, 18101, 16597, 19790, 19148, 18556, 19904, 18454, 21009, 20369, 21128, 20932, 18738, 21558, 18474, 21116, 19820, 17114, 19056, 16118, 17543, 17911, 15672, 17568, 17190, 16721, 14801, 16232, 16359, 14084, 15901, 16576, 14584, 16598, 15642, 17028, 17791, 15826, 17573, 16294, 17610, 18394, 18038, 18514, 17829, 19453, 17992, 20473, 20042, 18745, 19884, 17932, 20704, 19903, 18587, 19957, 19659, 20286, 17691, 20508, 20343, 18544, 21237, 20576, 18031, 20049, 16705, 18796, 17822, 15893, 18329, 15456, 19854, 15521, 17042, 16972, 14350, 15526, 14155, 15253, 15034, 12899, 14015, 13790, 11762, 13002, 11305, 12491, 13047, 12793, 11799, 13769, 13376, 12398, 13218, 12974, 11865, 12523, 11408, 12257, 12249, 11243, 11988, 12095, 11481, 10329, 11479, 11115, 9501, 10360, 8846, 9883, 9548, 8501, 9342, 8987, 8140, 9050, 7703, 8985, 8216, 8057, 7508, 7951, 8016, 6715, 7415, 7453, 6199, 6911, 5830, 6213, 6309, 5453, 5810, 5954, 5670, 5127, 5351, 5424, 4955, 5369, 5059, 5279, 5411, 5031, 5315, 5650, 4873, 5498, 5270, 5640, 5486, 5730, 5572, 4992, 5342, 4942, 5538, 5395, 4951, 5190, 4634, 5486, 5224, 4576, 5273, 5068, 5234, 4510, 4987, 5185, 4438, 4973, 4840, 4249, 4828, 4163, 4632, 4865, 4272, 4715, 4183, 4570, 4888, 4572, 4816, 4089, 4526, 4331, 4587, 4753, 4164, 4591, 4790, 4206, 4505, 4232, 4626, 112561, 0]],
+    ["8/333859634267203948016118549007441698534_uwr29s.png", [197384, 3762, 1369, 1334, 1395, 1595, 1926, 2249, 2532, 2644, 2897, 2468, 2378, 2035, 2259, 2175, 2494, 2334, 2562, 2585, 2871, 2968, 3296, 3247, 3240, 3256, 3410, 3361, 3281, 3514, 3651, 3588, 3806, 4145, 4640, 4984, 5552, 6320, 6891, 7834, 8292, 9131, 10701, 10698, 12503, 12915, 15505, 15069, 17255, 17475, 18302, 19504, 19210, 21324, 21976, 21845, 22810, 23324, 25640, 23114, 25154, 23772, 27263, 24559, 26156, 24730, 26205, 24818, 24750, 24299, 26058, 23064, 24756, 23025, 24695, 25177, 23767, 23694, 25906, 24191, 26519, 24552, 24639, 25223, 25472, 24044, 26070, 26329, 26674, 25620, 27633, 28035, 26682, 26914, 27654, 30294, 30761, 28871, 28868, 31199, 27313, 29155, 29075, 28828, 28954, 28719, 26139, 27992, 27505, 25392, 27397, 27167, 26840, 25118, 24682, 28713, 25122, 24943, 25295, 27756, 25600, 26463, 26850, 27024, 28139, 28463, 29407, 32209, 31149, 32024, 33300, 34006, 35035, 33281, 36901, 38055, 38863, 40109, 38213, 42411, 43246, 40821, 44952, 49059, 42409, 42841, 47471, 47406, 43626, 44150, 44470, 49013, 49121, 45831, 46707, 50751, 47165, 47459, 47735, 55790, 47982, 48246, 48090, 52350, 47494, 47202, 46504, 45508, 44375, 43414, 42542, 42194, 41426, 41280, 40815, 44417, 40793, 41086, 41150, 41719, 41787, 42668, 43055, 39754, 47335, 40783, 44309, 44699, 44983, 45223, 41487, 48810, 45196, 45483, 41849, 45760, 45817, 45715, 42312, 46541, 46199, 46831, 44218, 48151, 47783, 47317, 43379, 46502, 49218, 39854, 38440, 39564, 36323, 32955, 27414, 24583, 25731, 18628, 16098, 14629, 12159, 8972, 8024, 6479, 5241, 3799, 2835, 2591, 1959, 1435, 1156, 822, 683, 459, 389, 304, 191, 126, 105, 71, 33, 21, 12, 9, 6, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0]],
+    ["1/308671648074000163114167886553150666431_2pac7b.png", [415725, 13365, 13405, 13074, 13057, 12859, 12510, 11188, 9033, 7554, 7159, 7240, 7207, 7190, 7471, 7818, 8271, 7439, 7543, 7349, 7379, 6975, 6660, 6544, 6835, 7064, 7071, 7279, 7295, 7467, 7740, 7915, 8942, 8934, 8867, 9090, 9398, 9390, 9700, 9960, 10418, 10882, 10917, 11266, 11461, 11361, 11656, 11952, 13485, 13174, 14111, 14915, 15179, 16118, 17246, 18101, 18917, 20147, 21127, 22264, 23308, 24464, 25080, 26365, 29050, 28459, 29950, 30309, 31724, 32143, 33019, 33685, 34455, 35596, 35374, 36142, 36872, 37187, 37036, 37781, 40134, 37666, 38461, 38311, 35561, 38704, 38455, 38893, 41009, 38520, 38516, 38387, 37630, 36877, 36713, 35997, 35313, 34617, 34526, 33365, 33124, 33245, 32861, 33154, 35506, 32917, 33201, 33059, 32954, 33228, 33461, 33694, 33429, 33863, 33884, 33993, 33256, 34315, 33293, 34024, 36341, 33891, 34925, 34970, 35946, 35969, 36549, 36860, 37578, 38166, 38764, 39110, 39932, 40924, 41529, 45547, 42360, 43596, 43185, 43975, 44099, 44909, 45731, 46346, 47064, 47151, 47223, 47559, 47686, 48352, 48266, 51367, 50013, 50471, 50838, 52617, 53091, 55417, 55411, 57044, 59114, 60543, 61597, 62692, 62779, 65087, 65925, 69340, 65062, 61217, 63925, 63177, 61967, 60985, 57523, 61101, 55443, 54322, 52669, 52419, 51823, 50422, 51314, 50217, 50877, 49093, 49246, 49808, 48752, 48014, 47755, 49096, 46330, 45417, 44649, 44399, 43232, 43687, 43457, 42807, 44149, 43206, 44650, 45224, 45816, 47155, 47182, 51341, 48003, 47250, 46581, 46459, 44426, 42731, 40682, 38041, 35549, 33123, 30073, 27003, 23769, 21747, 18976, 17217, 13738, 11694, 9882, 8089, 6748, 5188, 4225, 3191, 2474, 1835, 1281, 953, 668, 469, 287, 214, 109, 68, 34, 12, 7, 3, 3, 1, 0, 0, 0, 0, 0, 0, 731, 0]],
+    ["18/216840111366964012810946289282010224084149790_02-122-189.png", [0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 230289, 18887, 17066, 41210, 30652, 32960, 20785, 18608, 15907, 16909, 15839, 10107, 21932, 15201, 11076, 13022, 11474, 12948, 11910, 5668, 13725, 12070, 6150, 13745, 8910, 9925, 10784, 9332, 6540, 3629, 18143, 3408, 15974, 3905, 10922, 7430, 13417, 4258, 8085, 11817, 13965, 4495, 8446, 12800, 9509, 5656, 9805, 4774, 15638, 11945, 7117, 19736, 7231, 16374, 30465, 0, 20772, 34906, 13476, 30847, 33750, 15981, 33631, 37916, 20491, 47676, 22205, 21555, 43515, 46025, 50788, 30304, 51577, 24318, 49358, 52913, 28530, 60081, 54803, 27246, 57173, 61877, 36206, 31595, 92530, 30931, 65762, 75760, 34673, 0, 67676, 66797, 34978, 72941, 79296, 36031, 68811, 34375, 72694, 38898, 82268, 73556, 36607, 75563, 82037, 90861, 42007, 41394, 82027, 88214, 47650, 102765, 93194, 46871, 47943, 104943, 58135, 127917, 111671, 111326, 58879, 127323, 201914, 62101, 123191, 130999, 69929, 151832, 129980, 62629, 192061, 68527, 213370, 63911, 62974, 62389, 202724, 81890, 204539, 128085, 133249, 152359, 68340, 129314, 181233, 60688, 194370, 59177, 163140, 107507, 119867, 108043, 99424, 47698, 146249, 104954, 90618, 84496, 128586, 92688, 80290, 113857, 78824, 85740, 110984, 72060, 74244, 116923, 69194, 68090, 72569, 78101, 69700, 140514, 38547, 119647, 106158, 112682, 46489, 110995, 99749, 68850, 107773, 87431, 55629, 89837, 27047, 71928, 46976, 76268, 43984, 81128, 67245, 39680, 76440, 20897, 65079, 57063, 63547, 48647, 61655, 41493, 72708, 86208, 19796, 64448, 89527, 58164, 42593, 68758, 57594, 61684, 68091, 56611, 219333, 191372, 222988, 135915, 315853, 162814, 340543, 339464, 136999, 17744, 0, 0, 0, 0, 0, 0, 0, 0]],
+    ["16/216840111366964012809176623042010216143036077_02-118-034.png", [466312, 1995, 1878, 2065, 2468, 2506, 2484, 2546, 2604, 2547, 2594, 2808, 2856, 3090, 3184, 3140, 3651, 3155, 3195, 3363, 3453, 3427, 3567, 3531, 3824, 3655, 3663, 3708, 3790, 3729, 3575, 3816, 4133, 3808, 4060, 4181, 4516, 4369, 4395, 4506, 4603, 4566, 4531, 4687, 4400, 4673, 4985, 5361, 6445, 5619, 5790, 6135, 6685, 7300, 7776, 9107, 9119, 9667, 10070, 10862, 11680, 12225, 13394, 13342, 15352, 14940, 16000, 17809, 17741, 18297, 18472, 18212, 18262, 17938, 18481, 17584, 17567, 17602, 17761, 17813, 20835, 18751, 18860, 18374, 17256, 16899, 17129, 16289, 16438, 16915, 17307, 17046, 17496, 18361, 17883, 18053, 20222, 18232, 19346, 18085, 18509, 18410, 18545, 18274, 18264, 19304, 18569, 18671, 18631, 18631, 18430, 18688, 18499, 19680, 21060, 19014, 19513, 20314, 19338, 18864, 18805, 18620, 18620, 18261, 18100, 19278, 18481, 18879, 18737, 18897, 19071, 18767, 19846, 19118, 21014, 19159, 19157, 19990, 19691, 20094, 20180, 20028, 20723, 20917, 21887, 20965, 21229, 21178, 21026, 21344, 21162, 21857, 21169, 20778, 22339, 20196, 20734, 19580, 18974, 18699, 18624, 17921, 17795, 18351, 17489, 17624, 17461, 17578, 17440, 17525, 17002, 17530, 16474, 15945, 17505, 16346, 15685, 15737, 15563, 15580, 15430, 15215, 15179, 15626, 14761, 14600, 14604, 14489, 14704, 14595, 15343, 16139, 15094, 15168, 15103, 15938, 15224, 14972, 14764, 14389, 13997, 13878, 14197, 13761, 13211, 13297, 13195, 15087, 13303, 13380, 13483, 13542, 14013, 13848, 14683, 14172, 14280, 14548, 14699, 14705, 14711, 15554, 14876, 16410, 14830, 14517, 15229, 14578, 14454, 14248, 14071, 13584, 13343, 13169, 12089, 11446, 10719, 9796, 9571, 9791, 8004, 7492, 7002, 6533, 6273, 6359, 5839, 5659, 5511, 5290, 5158, 5037, 4873, 4796, 131862, 0]]
+  ]
+}

lfs @ fe158e39

-Subproject commit 05344d20182ad4169fc5b9c38052d629aded30ed
+Subproject commit fe158e393b7904c6b381b0d9862895bf05c6cf7a