Merge branch 'develop' into dev-add-python3.14

shaneahmed · web-flow · commit abaa1729be68 · 2026-03-30T12:15:16.000+01:00
diff --git a/.github/workflows/docker-publish.yml b/.github/workflows/docker-publish.yml
@@ -6,7 +6,7 @@ env:
   REGISTRY: ghcr.io
   IMAGE_NAME: TissueImageAnalytics/tiatoolbox
   image: ghcr.io/tissueimageanalytics/tiatoolbox
-  TOOLBOX_VER: 2.0.0
+  TOOLBOX_VER: 2.0.1
 
 jobs:
   build-and-push-image:
diff --git a/AUTHORS.md b/AUTHORS.md
@@ -30,6 +30,7 @@
 - Fayyaz Minhas \<@foxtrotmike>
 - George Batchkala \<@GeorgeBatch>
 - Aleksandar Acic \<@aacic>
+- Esha Nasir \<@eshasadia>
 - Rob Jewsbury \<@R-J96>
 - Mohsin Bilal \<@mbhahsmi>
 - Musraf Basheer \<mbasheer04>
diff --git a/CITATION.cff b/CITATION.cff
@@ -44,7 +44,7 @@ authors:
   given-names: "Shan E Ahmed"
   orcid: "https://orcid.org/0000-0002-1097-1738"
 title: "TIAToolbox as an end-to-end library for advanced tissue image analytics"
-version: 2.0.0  # TIAToolbox version
+version: 2.0.1  # TIAToolbox version
 doi: 10.5281/zenodo.5802442
 date-released: 2022-10-20
 url: "https://github.com/TissueImageAnalytics/tiatoolbox"
diff --git a/HISTORY.md b/HISTORY.md
@@ -1,5 +1,16 @@
 # History
 
+## TIAToolbox v2.0.1 (2026-03-16)
+
+### Bug Fixes and Other Changes
+
+- Fixes filtering of tiles for post-processing within masked region (#1037).
+- Fixes Calculation of `max_inst_value` for Instances where overlap is empty (#1038)
+
+**Full Changelog:** https://github.com/TissueImageAnalytics/tiatoolbox/compare/v2.0.0...v2.0.1
+
+______________________________________________________________________
+
 ## TIAToolbox v2.0.0 (2026-03-11)
 
 ### ✨ Major Updates and Feature Improvements
@@ -58,6 +69,10 @@ TIAViz now integrates Meta’s Segment Anything Model (SAM), enabling:
 
 Simplified SAM usage (#968) streamlines its integration into analysis pipelines.
 
+### Multichannel Image (mIF) Support in WSIReader
+
+- **Multichannel image support** (PR #825) for immunofluorescence and non-RGB modalities
+
 ### 🖼️ WSI Registration Visualization in TIAViz
 
 TIAViz now supports **interactive WSI registration visualisation**, allowing users to compare aligned slides in two modes:
@@ -72,7 +87,6 @@ This feature enables intuitive, high‑resolution exploration of slide registrat
 Major improvements include:
 
 - More robust cross-vendor **metadata extraction** (#1001)
-- **Multichannel image support** (PR #825) for immunofluorescence and non-RGB modalities
 - Simplified Windows installation using `openslide-bin` (no manual DLL steps)
 - macOS Tileserver fix (#976)
 - Improved DICOM reading (#934)
diff --git a/docs/pretrained.rst b/docs/pretrained.rst
@@ -298,7 +298,7 @@ The input output configuration is as follows:
 
 .. collapse:: Model names
 
-    - micronet_hovernet-consep
+    - micronet-consep
 
 
 Kumar Dataset
@@ -333,7 +333,7 @@ The input output configuration is as follows:
 
 .. collapse:: Model names
 
-    - hovernet_original_kumar
+    - hovernet_original-kumar
 
 Nucleus Detection
 ^^^^^^^^^^^^^^^^^
diff --git a/pyproject.toml b/pyproject.toml
@@ -25,7 +25,7 @@
 
 [tool.poetry]
 name = "TIA Centre"
-version = "2.0.0"
+version = "2.0.1"
 description = "test"
 authors = ["TIA Centre <TIA@warwick.ac.uk>"]
 
diff --git a/requirements/requirements.txt b/requirements/requirements.txt
@@ -29,7 +29,7 @@ scikit-learn>=1.2.0
 scipy>=1.8
 shapely>=2.0.0
 SimpleITK>=2.2.1
-sphinx>=5.3.0
+sphinx>=5.3.0, <9.0.0
 tifffile>=2025.5.21
 timm>=1.0.3
 torch>=2.5.0
diff --git a/setup.py b/setup.py
@@ -63,6 +63,6 @@
     test_suite="tests",
     tests_require=test_requirements,
     url="https://github.com/TissueImageAnalytics/tiatoolbox",
-    version="2.0.0",
+    version="2.0.1",
     zip_safe=False,
 )
diff --git a/tests/engines/test_multi_task_segmentor.py b/tests/engines/test_multi_task_segmentor.py
@@ -5,6 +5,7 @@
 import json
 import shutil
 from pathlib import Path
+from types import SimpleNamespace
 from typing import TYPE_CHECKING, Any, Final
 from unittest.mock import MagicMock
 
@@ -1029,7 +1030,6 @@ def test_get_tile_info_small_image_triggers_early_return(
 
     """
     # --- Arrange ---
-    # Make image smaller than tile_shape
     image_shape = [100, 100]
 
     # Configure tile_shape so that tile_shape >= image_shape
@@ -1041,25 +1041,35 @@ def test_get_tile_info_small_image_triggers_early_return(
         patch_input_shape=(200, 200),
     )
 
-    # Patch PatchExtractor.get_coordinates to return predictable boxes
+    # Fake return from PatchExtractor.get_coordinates
     fake_boxes = np.array([[0, 0, 100, 100]])
     monkeypatch.setattr(
         "tiatoolbox.tools.patchextraction.PatchExtractor.get_coordinates",
-        lambda **kwargs: (None, fake_boxes),  # noqa: ARG005
+        lambda **_: (None, fake_boxes),
     )
 
+    # Create a dummy dataset with required attributes
+    dummy_dataset = SimpleNamespace(mask_reader=None)
+
+    # Dummy dataloader-like container
+    dummy_dataloader = SimpleNamespace(dataset=dummy_dataset)
+
+    # Create a real instance and inject required fields
+    m = MultiTaskSegmentor(model="hovernet_fast-pannuke")
+    m._ioconfig = ioconfig
+    m.mask_padding = (0, 0, 0, 0)
+    m.dataloader = dummy_dataloader
+
     # --- Act ---
-    result = MultiTaskSegmentor._get_tile_info(image_shape, ioconfig)
+    result = m._get_tile_info(image_shape=image_shape, wsi_proc_shape=image_shape)
 
     # --- Assert ---
     assert isinstance(result, list)
-    assert len(result) == 1  # early return path
+    assert len(result) == 1
+
     boxes, flag = result[0]
 
-    # boxes should be exactly what we patched
     assert np.array_equal(boxes, fake_boxes)
-
-    # flag should be zeros with shape (N, 4)
     assert flag.shape == (1, 4)
     assert np.all(flag == 0)
 
diff --git a/tests/models/test_arch_micronet.py b/tests/models/test_arch_micronet.py
@@ -72,7 +72,7 @@ def test_micronet_output(remote_sample: Callable, track_tmp_path: Path) -> None:
     svs_1_small = Path(remote_sample("svs-1-small"))
     micronet_output = Path(remote_sample("micronet-output"))
     model = "micronet-consep"
-    batch_size = 64
+    batch_size = 16
     num_workers = 0
 
     ninst_seg = NucleusInstanceSegmentor(
diff --git a/tiatoolbox/__init__.py b/tiatoolbox/__init__.py
@@ -16,7 +16,7 @@
 
 __author__ = """TIA Centre"""
 __email__ = "TIA@warwick.ac.uk"
-__version__ = "2.0.0"
+__version__ = "2.0.1"
 
 # This will set the tiatoolbox external data
 # default to be the user home folder, should work on both Window and Unix/Linux
diff --git a/tiatoolbox/models/dataset/dataset_abc.py b/tiatoolbox/models/dataset/dataset_abc.py
@@ -334,15 +334,15 @@ def __init__(  # skipcq: PY-R1000  # noqa: PLR0913
                 stride_shape=stride_shape[::-1],
             )
 
-        mask_reader = self._setup_mask_reader(
+        self.mask_reader = self._setup_mask_reader(
             input_mask=input_mask,
             reader=reader,
             auto_get_mask=auto_get_mask,
         )
 
-        if mask_reader is not None:
+        if self.mask_reader is not None:
             selected = PatchExtractor.filter_coordinates(
-                mask_reader,  # must be at the same resolution
+                self.mask_reader,  # must be at the same resolution
                 self.inputs,  # must already be at requested resolution
                 wsi_shape=wsi_shape,
                 min_mask_ratio=min_mask_ratio,
diff --git a/tiatoolbox/models/engine/multi_task_segmentor.py b/tiatoolbox/models/engine/multi_task_segmentor.py
@@ -1166,7 +1166,9 @@ def _process_tile_mode(
 
         # * retrieve tile placement and tile info flag
         # tile shape will always be corrected to be multiple of output
-        tile_info_sets = self._get_tile_info(masked_output_shape, self._ioconfig)
+        tile_info_sets = self._get_tile_info(
+            image_shape=masked_output_shape, wsi_proc_shape=wsi_proc_shape
+        )
         ioconfig = self._ioconfig.to_baseline()
 
         tile_metadata = _build_tile_tasks(
@@ -1358,10 +1360,10 @@ def _compute_tile(
         postproc_func = self._get_model_attr("postproc_func")
         return postproc_func(head_raws)  # offset is (0, 0) by default.
 
-    @staticmethod
     def _get_tile_info(
+        self: MultiTaskSegmentor,
         image_shape: list[int, int] | tuple[int, int] | np.ndarray,
-        ioconfig: IOSegmentorConfig,
+        wsi_proc_shape: tuple[int, int] | np.ndarray,
     ) -> list[list, ...]:
         """Generating tile information.
 
@@ -1379,8 +1381,8 @@ def _get_tile_info(
             image_shape (:class:`numpy.ndarray`, list(int, int), tuple(int, int)):
                 The shape of WSI to extract the tile from, assumed to be
                 in `[width, height]` / `[x, y]`.
-            ioconfig (:obj:IOSegmentorConfig):
-                The input and output configuration objects.
+            wsi_proc_shape (:class:`numpy.ndarray`, list(int, int), tuple(int, int)):
+                Full wsi processing shape.
 
         Returns:
             list:
@@ -1398,6 +1400,7 @@ def _get_tile_info(
                     - :class:`numpy.ndarray` - Removal flags
 
         """
+        ioconfig = self._ioconfig
         margin = 0 if ioconfig.margin is None else np.array(ioconfig.margin)
         tile_shape = np.array(ioconfig.tile_shape)
         tile_shape = (
@@ -1415,6 +1418,21 @@ def _get_tile_info(
         # * === Now generating the flags to indicate which side should
         # * === be removed in postproc callback
         boxes = tile_outputs[1]
+        offset = self.mask_padding[:2]
+        boxes = boxes + np.array((*offset, *offset))
+        mask_reader = self.dataloader.dataset.mask_reader
+
+        # Filter masked regions
+        if mask_reader is not None:
+            selected = PatchExtractor.filter_coordinates(
+                mask_reader,  # must be at the same resolution
+                boxes,  # must already be at requested resolution
+                wsi_shape=wsi_proc_shape,
+                min_mask_ratio=0,
+            )
+            boxes = boxes[selected]
+
+        boxes = boxes - np.array((*offset, *offset))
 
         # This saves computation time if the image is smaller than the expected tile
         if np.all(image_shape <= tile_shape):
@@ -1440,7 +1458,6 @@ def unset_removal_flag(boxes: tuple, removal_flag: np.ndarray) -> np.ndarray:
             return removal_flag
 
         w, h = image_shape
-        boxes = tile_outputs[1]
         #  expand to full four corners
         boxes_br = boxes[:, 2:]
         boxes_tr = np.dstack([boxes[:, 2], boxes[:, 1]])[0]
@@ -3183,7 +3200,11 @@ def _update_tile_based_predictions_array(
             final_combined = previous_predictions_.copy()
             final_combined[keep_mask] = new_predictions_[keep_mask] + max_inst_value
             new_predictions_ = final_combined  # Update for final merge
-            max_inst_value = np.max(new_predictions_[:][:]) + max_inst_value
+            max_inst_value = (
+                np.max(new_predictions_[:][:]) + max_inst_value
+                if np.any(keep_mask)
+                else max_inst_value
+            )
 
         wsi_info_dict[idx]["predictions"][y_start:y_end, x_start:x_end] = (
             new_predictions_

Original file line number	Diff line number	Diff line change
`@@ -63,6 +63,6 @@`
`63`	`63`	`test_suite="tests",`
`64`	`64`	`tests_require=test_requirements,`
`65`	`65`	`url="https://github.com/TissueImageAnalytics/tiatoolbox",`
`66`		`- version="2.0.0",`
	`66`	`+ version="2.0.1",`
`67`	`67`	`zip_safe=False,`
`68`	`68`	`)`