NEW: Add micronet

docs/pretrained.rst

@@ -113,7 +113,7 @@ They share the same input output configuration defined below:
             patch_input_shape=(1024, 1024),
             patch_output_shape=(512, 512),
             stride_shape=(256, 256),
-            save_resolution={'units': 'mpp', 'resolution': 8.0}        
+            save_resolution={'units': 'mpp', 'resolution': 8.0}
         )
 
 
@@ -144,7 +144,7 @@ They share the same input output configuration defined below:
             patch_input_shape=(1024, 1024),
             patch_output_shape=(512, 512),
             stride_shape=(256, 256),
-            save_resolution={'units': 'mpp', 'resolution': 0.25}        
+            save_resolution={'units': 'mpp', 'resolution': 0.25}
         )
 
 
@@ -183,7 +183,7 @@ input output configuration:
             patch_input_shape=(256, 256),
             patch_output_shape=(164, 164),
             stride_shape=(164, 164),
-            save_resolution={'units': 'mpp', 'resolution': 0.25}        
+            save_resolution={'units': 'mpp', 'resolution': 0.25}
         )
 
 .. collapse:: Model names
@@ -217,7 +217,7 @@ input output configuration:
             patch_input_shape=(256, 256),
             patch_output_shape=(164, 164),
             stride_shape=(164, 164),
-            save_resolution={'units': 'mpp', 'resolution': 0.25}        
+            save_resolution={'units': 'mpp', 'resolution': 0.25}
         )
 
 .. collapse:: Model names
@@ -251,12 +251,13 @@ input output configuration:
             patch_input_shape=(270, 270),
             patch_output_shape=(80, 80),
             stride_shape=(80, 80),
-            save_resolution={'units': 'mpp', 'resolution': 0.25}        
+            save_resolution={'units': 'mpp', 'resolution': 0.25}
         )
 
 .. collapse:: Model names
 
     - hovernet_original-consep
+    - micronet_hovernet-consep
 
 
 --------------------
@@ -285,7 +286,7 @@ input output configuration:
             patch_input_shape=(270, 270),
             patch_output_shape=(80, 80),
             stride_shape=(80, 80),
-            save_resolution={'units': 'mpp', 'resolution': 0.25}        
+            save_resolution={'units': 'mpp', 'resolution': 0.25}
         )
 
 .. collapse:: Model names
@@ -325,9 +326,9 @@ input output configuration:
             patch_input_shape=(256, 256),
             patch_output_shape=(164, 164),
             stride_shape=(164, 164),
-            save_resolution={'units': 'mpp', 'resolution': 0.5}        
+            save_resolution={'units': 'mpp', 'resolution': 0.5}
         )
 
 .. collapse:: Model names
 
-    - hovernetplus-oed
+    - hovernetplus-oed

docs/usage.rst

@@ -110,6 +110,7 @@ Neural Network Architectures
 - :obj:`Simplified U-Nets <tiatoolbox.models.architecture.unet>`
 - :obj:`HoVerNet <tiatoolbox.models.architecture.hovernet.HoVerNet>`
 - :obj:`HoVerNet+ <tiatoolbox.models.architecture.hovernetplus.HoVerNetPlus>`
+- :obj:`MicroNet <tiatoolbox.models.architecture.micronet.MicroNet>`
 
 Pipelines:
     - :obj:`IDARS <tiatoolbox.models.architecture.idars>`

tests/models/test_arch_micronet.py

@@ -0,0 +1,98 @@
+# ***** BEGIN GPL LICENSE BLOCK *****
+#
+# This program is free software; you can redistribute it and/or
+# modify it under the terms of the GNU General Public License
+# as published by the Free Software Foundation; either version 2
+# of the License, or (at your option) any later version.
+#
+# This program is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# GNU General Public License for more details.
+#
+# You should have received a copy of the GNU General Public License
+# along with this program; if not, write to the Free Software Foundation,
+# Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
+#
+# The Original Code is Copyright (C) 2021, TIA Centre, University of Warwick
+# All rights reserved.
+# ***** END GPL LICENSE BLOCK *****
+
+"""Unit test package for HoVerNet."""
+
+import pathlib
+
+import numpy as np
+import pytest
+import torch
+
+from tiatoolbox import utils
+from tiatoolbox.models.architecture import fetch_pretrained_weights
+from tiatoolbox.models.architecture.micronet import MicroNet
+from tiatoolbox.models.engine.semantic_segmentor import SemanticSegmentor
+from tiatoolbox.utils import env_detection as toolbox_env
+from tiatoolbox.wsicore.wsireader import WSIReader
+
+
+def test_functionality(remote_sample, tmp_path):
+    """Functionality test."""
+    tmp_path = str(tmp_path)
+    sample_wsi = str(remote_sample("wsi1_2k_2k_svs"))
+    reader = WSIReader.open(sample_wsi)
+
+    # * test fast mode (architecture used in PanNuke paper)
+    patch = reader.read_bounds(
+        (0, 0, 252, 252), resolution=0.25, units="mpp", coord_space="resolution"
+    )
+    patch = MicroNet.preproc(patch)
+    batch = torch.from_numpy(patch)[None]
+    model = MicroNet()
+    fetch_pretrained_weights("micronet_hovernet-consep", f"{tmp_path}/weights.pth")
+    map_location = utils.misc.select_device(utils.env_detection.has_gpu())
+    pretrained = torch.load(f"{tmp_path}/weights.pth", map_location=map_location)
+    model.load_state_dict(pretrained)
+    output = model.infer_batch(model, batch, on_gpu=False)
+    output, _ = model.postproc(output[0])
+    assert np.max(np.unique(output)) == 33
+
+
+def test_value_error():
+    """Test to generate value error is num_classes < 2."""
+    with pytest.raises(ValueError, match="Number of classes should be >=2"):
+        _ = MicroNet(num_class=1)
+
+
+@pytest.mark.skipif(
+    toolbox_env.running_on_travis() or not toolbox_env.has_gpu(),
+    reason="Local test on machine with GPU.",
+)
+def test_micronet_output(remote_sample, tmp_path):
+    """Tests the output of MicroNet."""
+    svs_1_small = pathlib.Path(remote_sample("svs-1-small"))
+    micronet_output = pathlib.Path(remote_sample("micronet-output"))
+    pretrained_model = "micronet_hovernet-consep"
+    batch_size = 5
+    num_loader_workers = 0
+    num_postproc_workers = 0
+
+    predictor = SemanticSegmentor(
+        pretrained_model=pretrained_model,
+        batch_size=batch_size,
+        num_loader_workers=num_loader_workers,
+        num_postproc_workers=num_postproc_workers,
+    )
+
+    output = predictor.predict(
+        imgs=[
+            svs_1_small,
+        ],
+        save_dir=tmp_path / "output",
+    )
+
+    output = np.load(output[0][1] + ".raw.0.npy")
+    output_on_server = np.load(str(micronet_output))
+    output_on_server = np.round(output_on_server, decimals=3)
+    new_output = np.round(output[500:1000, 1000:1500, :], decimals=3)
+    true_values = output_on_server == new_output
+    percent_true = np.count_nonzero(true_values) / np.size(output_on_server)
+    assert percent_true > 0.999

tests/models/test_hovernet.py

@@ -32,23 +32,23 @@
     ResidualBlock,
     TFSamepaddingLayer,
 )
-from tiatoolbox.wsicore.wsireader import get_wsireader
+from tiatoolbox.wsicore.wsireader import WSIReader
 
 
 def test_functionality(remote_sample, tmp_path):
     """Functionality test."""
     tmp_path = str(tmp_path)
     sample_wsi = str(remote_sample("wsi1_2k_2k_svs"))
-    reader = get_wsireader(sample_wsi)
+    reader = WSIReader.open(sample_wsi)
 
-    # * test fast mode (architecture used in Pannuke paper)
+    # * test fast mode (architecture used in PanNuke paper)
     patch = reader.read_bounds(
-        [0, 0, 256, 256], resolution=0.25, units="mpp", coord_space="resolution"
+        (0, 0, 256, 256), resolution=0.25, units="mpp", coord_space="resolution"
     )
     batch = torch.from_numpy(patch)[None]
     model = HoVerNet(num_types=6, mode="fast")
-    fetch_pretrained_weights("hovernet_fast-pannuke", f"{tmp_path}/weigths.pth")
-    pretrained = torch.load(f"{tmp_path}/weigths.pth")
+    fetch_pretrained_weights("hovernet_fast-pannuke", f"{tmp_path}/weights.pth")
+    pretrained = torch.load(f"{tmp_path}/weights.pth")
     model.load_state_dict(pretrained)
     output = model.infer_batch(model, batch, on_gpu=False)
     output = [v[0] for v in output]
@@ -57,12 +57,12 @@ def test_functionality(remote_sample, tmp_path):
 
     # * test fast mode (architecture used for MoNuSAC data)
     patch = reader.read_bounds(
-        [0, 0, 256, 256], resolution=0.25, units="mpp", coord_space="resolution"
+        (0, 0, 256, 256), resolution=0.25, units="mpp", coord_space="resolution"
     )
     batch = torch.from_numpy(patch)[None]
     model = HoVerNet(num_types=5, mode="fast")
-    fetch_pretrained_weights("hovernet_fast-monusac", f"{tmp_path}/weigths.pth")
-    pretrained = torch.load(f"{tmp_path}/weigths.pth")
+    fetch_pretrained_weights("hovernet_fast-monusac", f"{tmp_path}/weights.pth")
+    pretrained = torch.load(f"{tmp_path}/weights.pth")
     model.load_state_dict(pretrained)
     output = model.infer_batch(model, batch, on_gpu=False)
     output = [v[0] for v in output]
@@ -71,12 +71,12 @@ def test_functionality(remote_sample, tmp_path):
 
     # * test original mode on CoNSeP dataset (architecture used in HoVerNet paper)
     patch = reader.read_bounds(
-        [0, 0, 270, 270], resolution=0.25, units="mpp", coord_space="resolution"
+        (0, 0, 270, 270), resolution=0.25, units="mpp", coord_space="resolution"
     )
     batch = torch.from_numpy(patch)[None]
     model = HoVerNet(num_types=5, mode="original")
-    fetch_pretrained_weights("hovernet_original-consep", f"{tmp_path}/weigths.pth")
-    pretrained = torch.load(f"{tmp_path}/weigths.pth")
+    fetch_pretrained_weights("hovernet_original-consep", f"{tmp_path}/weights.pth")
+    pretrained = torch.load(f"{tmp_path}/weights.pth")
     model.load_state_dict(pretrained)
     output = model.infer_batch(model, batch, on_gpu=False)
     output = [v[0] for v in output]
@@ -85,12 +85,12 @@ def test_functionality(remote_sample, tmp_path):
 
     # * test original mode on Kumar dataset (architecture used in HoVerNet paper)
     patch = reader.read_bounds(
-        [0, 0, 270, 270], resolution=0.25, units="mpp", coord_space="resolution"
+        (0, 0, 270, 270), resolution=0.25, units="mpp", coord_space="resolution"
     )
     batch = torch.from_numpy(patch)[None]
     model = HoVerNet(num_types=None, mode="original")
-    fetch_pretrained_weights("hovernet_original-kumar", f"{tmp_path}/weigths.pth")
-    pretrained = torch.load(f"{tmp_path}/weigths.pth")
+    fetch_pretrained_weights("hovernet_original-kumar", f"{tmp_path}/weights.pth")
+    pretrained = torch.load(f"{tmp_path}/weights.pth")
     model.load_state_dict(pretrained)
     output = model.infer_batch(model, batch, on_gpu=False)
     output = [v[0] for v in output]

tiatoolbox/data/pretrained_model.yaml

@@ -621,7 +621,7 @@ hovernet_fast-pannuke:
     ioconfig:
         class: semantic_segmentor.IOSegmentorConfig
         kwargs:
-            input_resolutions: 
+            input_resolutions:
                 - {"units": "mpp", "resolution": 0.25}
             output_resolutions:
                 - {"units": "mpp", "resolution": 0.25}
@@ -644,7 +644,7 @@ hovernet_fast-monusac:
     ioconfig:
         class: semantic_segmentor.IOSegmentorConfig
         kwargs:
-            input_resolutions: 
+            input_resolutions:
                 - {"units": "mpp", "resolution": 0.25}
             output_resolutions:
                 - {"units": "mpp", "resolution": 0.25}
@@ -667,7 +667,7 @@ hovernet_original-consep:
     ioconfig:
         class: semantic_segmentor.IOSegmentorConfig
         kwargs:
-            input_resolutions: 
+            input_resolutions:
                 - {"units": "mpp", "resolution": 0.25}
             output_resolutions:
                 - {"units": "mpp", "resolution": 0.25}
@@ -690,7 +690,7 @@ hovernet_original-kumar:
     ioconfig:
         class: semantic_segmentor.IOSegmentorConfig
         kwargs:
-            input_resolutions: 
+            input_resolutions:
                 - {"units": "mpp", "resolution": 0.25}
             output_resolutions:
                 - {"units": "mpp", "resolution": 0.25}
@@ -712,16 +712,37 @@ hovernetplus-oed:
     ioconfig:
         class: semantic_segmentor.IOSegmentorConfig
         kwargs:
-            input_resolutions: 
+            input_resolutions:
                 - {"units": "mpp", "resolution": 0.50}
             output_resolutions:
                 - {"units": "mpp", "resolution": 0.50}
                 - {"units": "mpp", "resolution": 0.50}
                 - {"units": "mpp", "resolution": 0.50}
-                - {"units": "mpp", "resolution": 0.50}    
+                - {"units": "mpp", "resolution": 0.50}
             margin: 128
             tile_shape: [2048, 2048]
             patch_input_shape: [256, 256]
             patch_output_shape: [164, 164]
             stride_shape: [164, 164]
             save_resolution: {'units': 'mpp', 'resolution': 0.50}
+
+micronet_hovernet-consep:
+    url: https://tiatoolbox.dcs.warwick.ac.uk/models/seg/micronet_hovernet-consep.pth
+    architecture:
+        class: micronet.MicroNet
+        kwargs:
+            num_input_channels: 3
+            num_class: 2
+    ioconfig:
+        class: semantic_segmentor.IOSegmentorConfig
+        kwargs:
+            input_resolutions:
+                - {"units": "mpp", "resolution": 0.25}
+            output_resolutions:
+                - {"units": "mpp", "resolution": 0.25}
+            margin: 128
+            tile_shape: [2048, 2048]
+            patch_input_shape: [252, 252]
+            patch_output_shape: [252, 252]
+            stride_shape: [150, 150]
+            save_resolution: {'units': 'mpp', 'resolution': 0.25}

tiatoolbox/data/remote_samples.yaml

@@ -98,3 +98,5 @@ files:
     url: [*modelroot, "predictions/bcss/wsi4_4k_4k.mask.npy"]
   small_svs_tissue_mask:
     url: [*modelroot, "predictions/tissue_mask/small_svs_tissue_mask.npy"]
+  micronet-output:
+    url: [*modelroot, "predictions/nuclei_mask/micronet-output.npy"]

tiatoolbox/models/architecture/__init__.py

@@ -32,7 +32,7 @@
 from tiatoolbox.models.dataset.classification import predefined_preproc_func
 from tiatoolbox.utils.misc import download_data
 
-__all__ = ["get_pretrained_model"]
+__all__ = ["get_pretrained_model", "fetch_pretrained_weights"]
 PRETRAINED_INFO = rcParam["pretrained_model_info"]
 
 

tiatoolbox/models/architecture/hovernet.py

@@ -559,7 +559,7 @@ def _proc_np_hv(np_map: np.ndarray, hv_map: np.ndarray, fx: float = 1):
         return proced_pred
 
     @staticmethod
-    def _get_instance_info(pred_inst, pred_type=None):
+    def get_instance_info(pred_inst, pred_type=None):
         """To collect instance information and store it within a dictionary.
 
         Args:
@@ -703,7 +703,7 @@ def postproc(raw_maps: List[np.ndarray]):
 
         pred_type = tp_map
         pred_inst = HoVerNet._proc_np_hv(np_map, hv_map)
-        nuc_inst_info_dict = HoVerNet._get_instance_info(pred_inst, pred_type)
+        nuc_inst_info_dict = HoVerNet.get_instance_info(pred_inst, pred_type)
 
         return pred_inst, nuc_inst_info_dict