[skip travis] DOC: Update docs for pretrained models

docs/pretrained.rst

@@ -1,7 +1,7 @@
 .. _pretrained-info-page:
 
 =================================
-Pre-trained Neural Network Models
+Pretrained Neural Network Models
 =================================
 
 ^^^^^^^^^^^^^^^^^^^^^^
@@ -29,8 +29,62 @@ They share the same input output configuration defined below:
 
 .. collapse:: Model names
 
-    - googlenet-kather100k
+    - alexnet-kather100k
+    - resnet18-kather100k
+    - resnet34-kather100k
+    - resnet50-kather100k
+    - resnet101-kather100k
+    - resnext50_32x4d-kather100k
+    - resnext101_32x8d-kather100k
+    - wide_resnet50_2-kather100k
+    - wide_resnet101_2-kather100k
+    - densenet121-kather100k
+    - densenet161-kather100k
+    - densenet169-kather100k
+    - densenet201-kather100k
+    - mobilenet_v2-kather100k
+    - mobilenet_v3_large-kather100k
     - mobilenet_v3_small-kather100k
+    - googlenet-kather100k
+
+-----------------------------
+Patch Camelyon (PCam) Dataset
+-----------------------------
+
+The following models are trained using the `PCam dataset <https://github.com/basveeling/pcam/>`_.
+They share the same input output configuration defined below:
+
+.. collapse:: Input Output Configuration Details
+
+   .. code-block:: python
+
+        from tiatoolbox.models import IOPatchPredictorConfig
+        ioconfig = IOPatchPredictorConfig(
+            patch_input_shape=(96, 96),
+            stride_shape=(96, 96),
+            input_resolutions=[{"resolution": 1.0, "units": "mpp"}]
+        )
+
+
+.. collapse:: Model names
+
+    - alexnet-pcam
+    - resnet18-pcam
+    - resnet34-pcam
+    - resnet50-pcam
+    - resnet101-pcam
+    - resnext50-pcam
+    - resnext101-pcam
+    - wide_resnet50_2-pcam
+    - wide_resnet101_2-pcam
+    - densenet121-pcam
+    - densenet161-pcam
+    - densenet169-pcam
+    - densenet201-pcam
+    - mobilenet_v2-pcam
+    - mobilenet_v3_large-pcam
+    - mobilenet_v3_small-pcam
+    - googlenet-pcam
 
 
 ^^^^^^^^^^^^^^^^^^^^^^
@@ -41,7 +95,7 @@ Semantic Segmentation
 Tissue Masking
 --------------------
 
-The following models are trained using internal data of TIACentre.
+The following models are trained using internal data of TIA Centre.
 They share the same input output configuration defined below:
 
 .. collapse:: Input Output Configuration Details
@@ -72,7 +126,7 @@ They share the same input output configuration defined below:
 Breast Cancer
 --------------------
 
-The following models are trained using `BCSS dataset <https://bcsegmentation.grand-challenge.org/>`_.
+The following models are trained using the `BCSS dataset <https://bcsegmentation.grand-challenge.org/>`_.
 They share the same input output configuration defined below:
 
 .. collapse:: Input Output Configuration Details
@@ -98,3 +152,182 @@ They share the same input output configuration defined below:
 
     - fcn_resnet50_unet-bcss
 
+
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+Nucleus Instance Segmentation
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+--------------------
+PanNuke Dataset
+--------------------
+
+We provide the following models trained using the `PanNuke dataset <https://warwick.ac.uk/fac/cross_fac/tia/data/pannuke>`_, which uses the following
+input output configuration:
+
+.. collapse:: Input Output Configuration Details
+
+   .. code-block:: python
+
+        from tiatoolbox.models import IOSegmentorConfig
+        ioconfig = IOSegmentorConfig(
+            input_resolutions=[
+                {'units': 'mpp', 'resolution': 0.25}
+            ],
+            output_resolutions=[
+                {'units': 'mpp', 'resolution': 0.25},
+                {'units': 'mpp', 'resolution': 0.25},
+                {'units': 'mpp', 'resolution': 0.25}
+            ],
+            margin=128
+            tile_shape=[1024, 1024]
+            patch_input_shape=(256, 256),
+            patch_output_shape=(164, 164),
+            stride_shape=(164, 164),
+            save_resolution={'units': 'mpp', 'resolution': 0.25}        
+        )
+
+.. collapse:: Model names
+
+    - hovernet_fast-pannuke
+
+
+--------------------
+MoNuSAC Dataset
+--------------------
+
+We provide the following models trained using the `MoNuSAC dataset <https://monusac.grand-challenge.org/>`_, which uses the following
+input output configuration:
+
+.. collapse:: Input Output Configuration Details
+
+   .. code-block:: python
+
+        from tiatoolbox.models import IOSegmentorConfig
+        ioconfig = IOSegmentorConfig(
+            input_resolutions=[
+                {'units': 'mpp', 'resolution': 0.25}
+            ],
+            output_resolutions=[
+                {'units': 'mpp', 'resolution': 0.25},
+                {'units': 'mpp', 'resolution': 0.25},
+                {'units': 'mpp', 'resolution': 0.25}
+            ],
+            margin=128
+            tile_shape=[1024, 1024]
+            patch_input_shape=(256, 256),
+            patch_output_shape=(164, 164),
+            stride_shape=(164, 164),
+            save_resolution={'units': 'mpp', 'resolution': 0.25}        
+        )
+
+.. collapse:: Model names
+
+    - hovernet_fast-monusac
+
+
+--------------------
+CoNSeP Dataset
+--------------------
+
+We provide the following models trained using the `CoNSeP dataset <https://warwick.ac.uk/fac/cross_fac/tia/data/hovernet/>`_, which uses the following
+input output configuration:
+
+.. collapse:: Input Output Configuration Details
+
+   .. code-block:: python
+
+        from tiatoolbox.models import IOSegmentorConfig
+        ioconfig = IOSegmentorConfig(
+            input_resolutions=[
+                {'units': 'mpp', 'resolution': 0.25}
+            ],
+            output_resolutions=[
+                {'units': 'mpp', 'resolution': 0.25},
+                {'units': 'mpp', 'resolution': 0.25},
+                {'units': 'mpp', 'resolution': 0.25}
+            ],
+            margin=128
+            tile_shape=[1024, 1024]
+            patch_input_shape=(270, 270),
+            patch_output_shape=(80, 80),
+            stride_shape=(80, 80),
+            save_resolution={'units': 'mpp', 'resolution': 0.25}        
+        )
+
+.. collapse:: Model names
+
+    - hovernet_original-consep
+
+
+--------------------
+Kumar Dataset
+--------------------
+
+We provide the following models trained using the `Kumar dataset <https://monuseg.grand-challenge.org/>`_, which uses the following
+input output configuration:
+
+.. collapse:: Input Output Configuration Details
+
+   .. code-block:: python
+
+        from tiatoolbox.models import IOSegmentorConfig
+        ioconfig = IOSegmentorConfig(
+            input_resolutions=[
+                {'units': 'mpp', 'resolution': 0.25}
+            ],
+            output_resolutions=[
+                {'units': 'mpp', 'resolution': 0.25},
+                {'units': 'mpp', 'resolution': 0.25},
+                {'units': 'mpp', 'resolution': 0.25}
+            ],
+            margin=128
+            tile_shape=[1024, 1024]
+            patch_input_shape=(270, 270),
+            patch_output_shape=(80, 80),
+            stride_shape=(80, 80),
+            save_resolution={'units': 'mpp', 'resolution': 0.25}        
+        )
+
+.. collapse:: Model names
+
+    - hovernet_original_kumar
+
+
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+Multi-Task Segmentation
+^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+
+----------------------------------------
+Oral Epithelial Dysplasia (OED) Dataset
+---------------------------------------
+
+We provide the following model trained using a private OED dataset. The model outputs nuclear instance segmentation
+and classification results, as well as semantic segmentation of epithelial layers. The model uses the following
+input output configuration:
+
+.. collapse:: Input Output Configuration Details
+
+   .. code-block:: python
+
+        from tiatoolbox.models import IOSegmentorConfig
+        ioconfig = IOSegmentorConfig(
+            input_resolutions=[
+                {'units': 'mpp', 'resolution': 0.5}
+            ],
+            output_resolutions=[
+                {'units': 'mpp', 'resolution': 0.5},
+                {'units': 'mpp', 'resolution': 0.5},
+                {'units': 'mpp', 'resolution': 0.5},
+                {'units': 'mpp', 'resolution': 0.5}
+            ],
+            margin=128
+            tile_shape=[1024, 1024]
+            patch_input_shape=(256, 256),
+            patch_output_shape=(164, 164),
+            stride_shape=(164, 164),
+            save_resolution={'units': 'mpp', 'resolution': 0.5}        
+        )
+
+.. collapse:: Model names
+
+    - hovernetplus-oed

tiatoolbox/models/engine/nucleus_instance_segmentor.py

@@ -282,7 +282,8 @@ class NucleusInstanceSegmentor(SemanticSegmentor):
           weights already loaded. Default is `None`. If provided,
           `pretrained_model` argument is ignored.
         pretrained_model (str): Name of the existing models support by tiatoolbox
-          for processing the data. Refer to [URL] for details.
+          for processing the data. For a full list of pretrained models, refer to the
+          `docs <https://tia-toolbox.readthedocs.io/en/latest/pretrained.html>`_.
           By default, the corresponding pretrained weights will also be
           downloaded. However, you can override with your own set of weights
           via the `pretrained_weights` argument. Argument is case insensitive.

tiatoolbox/models/engine/patch_predictor.py

@@ -79,6 +79,8 @@ class PatchPredictor:
           >>> predictor = PatchPredictor(
           ...    pretrained_model="resnet18-kather100k",
           ...    pretrained_weights="resnet18_local_weight")
+          For a full list of pretrained models, refer to the
+          `docs <https://tia-toolbox.readthedocs.io/en/latest/pretrained.html>`_
         batch_size (int) : Number of images fed into the model each time.
         num_loader_workers (int) : Number of workers to load the data.
           Take note that they will also perform preprocessing.
@@ -91,8 +93,8 @@ class PatchPredictor:
           or `wsi`.
         model (nn.Module): Defined PyTorch model.
         pretrained_model (str): Name of the existing models support by tiatoolbox
-          for processing the data. Refer to
-          `tiatoolbox.models.classification.get_pretrained_model` for details.
+          for processing the data. For a full list of pretrained models, refer to the
+          `docs <https://tia-toolbox.readthedocs.io/en/latest/pretrained.html>`_
           By default, the corresponding pretrained weights will also be
           downloaded. However, you can override with your own set of weights
           via the `pretrained_weights` argument. Argument is case insensitive.

tiatoolbox/models/engine/semantic_segmentor.py

@@ -369,7 +369,8 @@ class SemanticSegmentor:
           weights already loaded. Default is `None`. If provided,
           `pretrained_model` argument is ignored.
         pretrained_model (str): Name of the existing models support by tiatoolbox
-          for processing the data. Refer to [URL] for details.
+          for processing the data. For a full list of pretrained models, refer to the
+          `docs <https://tia-toolbox.readthedocs.io/en/latest/pretrained.html>`_.
           By default, the corresponding pretrained weights will also be
           downloaded. However, you can override with your own set of weights
           via the `pretrained_weights` argument. Argument is case insensitive.