Updated DeepLab guide with PointRend

Author: kapil-varshney

guide/14-deep-learning/how_deeplabv3_works.ipynb

@@ -4,14 +4,14 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "# How DeepLabV3 Works"
+    "## How DeepLabV3 Works"
    ]
   },
   {
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "## Introduction"
+    "### Introduction"
    ]
   },
   {
@@ -83,7 +83,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "## Atrous Convoltion (Dilated Convolution)"
+    "### Atrous Convoltion (Dilated Convolution)"
    ]
   },
   {
@@ -145,7 +145,7 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "## Atrous Spatial Pyramid Pooling (ASPP)"
+    "### Atrous Spatial Pyramid Pooling (ASPP)"
    ]
   },
   {
@@ -159,7 +159,34 @@
    "cell_type": "markdown",
    "metadata": {},
    "source": [
-    "### References:"
+    "### PointRend Enhancement\n",
+    "\n",
+    "Segmentation models can tend to over-smoothen boundaries which might not be precise for objects or scenes with irregular boundaries. To get a crisp segmentation boundary, a point-based rendering neural network module called [**PointRend**](https://arxiv.org/abs/1912.08193) has been added as an enhancement to the existing model. This module draws methodology from classical computer graphics and gives the perspective of rendering to a segmentation problem. An iterative subdivision algorithm at selected locations is used to make point-based segmentation predictions. This method enables high-resolution output in an efficient way. [8]"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "<center><img src=\"../../static/img/pointrend_deeplabv3.jpg\"/></center>\n",
+    "<center>Figure 4. PointRend enhancement (right) over original segmentation model (left) [8]</center>"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "To enable PointRend with DeepLabV3, initialize the model with parameter `pointrend=True`:\n",
+    "```\n",
+    "model = DeepLab(data=data, pointrend=True)\n",
+    "```"
+   ]
+  },
+  {
+   "cell_type": "markdown",
+   "metadata": {},
+   "source": [
+    "## References:"
    ]
   },
   {
@@ -183,8 +210,17 @@
     "[6] Sik-Ho Tsang, Review: DeepLabv3 — Atrous Convolution (Semantic Segmentation), https://towardsdatascience.com/review-deeplabv3-atrous-convolution-semantic-segmentation-6d818bfd1d74, Accessed 21 Februrary 2020\n",
     "\n",
     "\n",
-    "[7] Saurabh Pal, Semantic Segmentation: Introduction to the Deep Learning Technique Behind Google Pixel’s Camera!, https://www.analyticsvidhya.com/blog/2019/02/tutorial-semantic-segmentation-google-deeplab/, Accessed 21 February 2020\n"
+    "[7] Saurabh Pal, Semantic Segmentation: Introduction to the Deep Learning Technique Behind Google Pixel’s Camera!, https://www.analyticsvidhya.com/blog/2019/02/tutorial-semantic-segmentation-google-deeplab/, Accessed 21 February 2020\n",
+    "\n",
+    "[8] Alexander Kirillov, Yuxin Wu, Kaiming He, Ross Girshick: “PointRend: Image Segmentation as Rendering”, 2019; [http://arxiv.org/abs/1912.08193 arXiv:1912.08193].\n"
    ]
+  },
+  {
+   "cell_type": "code",
+   "execution_count": null,
+   "metadata": {},
+   "outputs": [],
+   "source": []
   }
  ],
  "metadata": {
@@ -203,7 +239,7 @@
    "name": "python",
    "nbconvert_exporter": "python",
    "pygments_lexer": "ipython3",
-   "version": "3.6.9"
+   "version": "3.7.9"
   }
  },
  "nbformat": 4,