Infer process needs more work

Signed-off-by: vnath <[email protected]>

Author: finalelement

nuclick/nuclick_infer.ipynb

@@ -0,0 +1,32 @@
+{
+ "cells": [
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
+  }
+ ],
+ "metadata": {
+  "kernelspec": {
+   "display_name": "nuclick_tutorial",
+   "language": "python",
+   "name": "nuclick_tutorial"
+  },
+  "language_info": {
+   "codemirror_mode": {
+    "name": "ipython",
+    "version": 3
+   },
+   "file_extension": ".py",
+   "mimetype": "text/x-python",
+   "name": "python",
+   "nbconvert_exporter": "python",
+   "pygments_lexer": "ipython3",
+   "version": "3.8.13"
+  }
+ },
+ "nbformat": 4,
+ "nbformat_minor": 4
+}

nuclick/nuclick_infer_process.py

@@ -0,0 +1,71 @@
+
+import logging
+from typing import Any, Callable, Dict, Sequence
+
+import numpy as np
+from monai.apps.nuclick.transforms import AddClickSignalsd, PostFilterLabeld
+from monai.transforms import Activationsd, AsChannelFirstd, AsDiscreted, EnsureTyped, SqueezeDimd, ToNumpyd, LoadImaged
+
+from monailabel.interfaces.tasks.infer import InferTask, InferType
+from monailabel.transform.post import FindContoursd
+from monailabel.transform.writer import PolygonWriter
+
+class NuClick(InferTask):
+    """
+    This provides Inference Engine for pre-trained NuClick segmentation (UNet) model.
+    """
+
+    def __init__(
+        self,
+        path,
+        network=None,
+        roi_size=(128, 128),
+        type=InferType.OTHERS,
+        labels=None,
+        dimension=2,
+        description="A pre-trained NuClick model for interactive cell segmentation for Pathology",
+        **kwargs,
+    ):
+        super().__init__(
+            path=path,
+            network=network,
+            roi_size=roi_size,
+            type=type,
+            labels=labels,
+            dimension=dimension,
+            description=description,
+            **kwargs,
+        )
+
+    def info(self) -> Dict[str, Any]:
+        d = super().info()
+        d["pathology"] = True
+        d["nuclick"] = True
+        return d
+
+    def pre_transforms(self, data=None):
+        return [
+            #LoadImagePatchd(keys="image", mode="RGB", dtype=np.uint8, padding=False),
+            LoadImaged(keys="image", mode="RGB", dtype=np.uint8, padding=False),
+            AsChannelFirstd(keys="image"),
+            AddClickSignalsd(image="image"),
+            EnsureTyped(keys="image", device=data.get("device") if data else None),
+        ]
+
+    def run_inferer(self, data, convert_to_batch=True, device="cuda"):
+        return super().run_inferer(data, False, device)
+
+    def post_transforms(self, data=None) -> Sequence[Callable]:
+        return [
+            EnsureTyped(keys="pred", device=data.get("device") if data else None),
+            Activationsd(keys="pred", sigmoid=True),
+            AsDiscreted(keys="pred", threshold=0.5),
+            SqueezeDimd(keys="pred", dim=1),
+            ToNumpyd(keys=("image", "pred")),
+            PostFilterLabeld(keys="pred"),
+            FindContoursd(keys="pred", labels=self.labels),
+        ]
+
+    def writer(self, data, extension=None, dtype=None):
+        writer = PolygonWriter(label=self.output_label_key, json=self.output_json_key)
+        return writer(data)

nuclick/nuclick_training_notebook.ipynb

@@ -8,6 +8,15 @@
     "\n",
     "This tutorial uses a deep learning framework for interactive segmentation for microscopic images, the framework is referred to as NuClick [1]\n",
     "\n",
+    "Add a figure from the paper, it always helps\n",
+    "\n",
+    "What features are included in this tutorial?\n",
+    "\n",
+    "From where to download the dataset\n",
+    "\n",
+    "Description of the dataset\n",
+    "\n",
+    "\n",
     "References:\n",
     "\n",
     "[1] Koohbanani, Navid Alemi, et al. \"NuClick: a deep learning framework for interactive segmentation of microscopic images.\" Medical Image Analysis 65 (2020): 101771."
@@ -1302,7 +1311,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": null,
+   "execution_count": 17,
    "metadata": {},
    "outputs": [
     {