minor notebook improvements

Author: Matt Sokoloff

examples/model_assisted_labeling/tiled_imagery_mal.ipynb

@@ -220,7 +220,9 @@
    "id": "portable-grenada",
    "metadata": {},
    "source": [
-    "#### Grab featureSchemaIds"
+    "* <b>Grab featureSchemaIds</b>\n",
+    "* This enables us to assign model inferences to a specific class in labelbox\n",
+    "* Checkout the ndjson section to see where this is used"
    ]
   },
   {
@@ -239,7 +241,7 @@
    ],
    "source": [
     "# When we created a project with the ontology defined above, all of the ids were assigned.\n",
-    "# So lets reconstruct the ontology builder with all of the ids.\n",
+    "# Let's reconstruct the ontology builder with all of the ids.\n",
     "ontology = ontology_builder.from_project(project)\n",
     "# We want all of the feature schemas to be easily accessible by name.\n",
     "schema_lookup = {tool.name: tool.feature_schema_id for tool in ontology.tools}\n",
@@ -252,18 +254,18 @@
    "metadata": {},
    "source": [
     "### Create import objects\n",
-    "* We want to create a json payload that matches this: https://docs.labelbox.com/en/model-assisted-labeling/create-import-file#images-7961\n",
-    "* Here we will run inferences on all of our data (only one image this time)\n",
-    "* See other notebook on supported objects. Everything is the same except coords are in lat lng"
+    "* We want to create a polygon json payload that matches the format documented here : https://docs.labelbox.com/data-model/en/index-en#polygon"
    ]
   },
   {
    "cell_type": "code",
-   "execution_count": 11,
+   "execution_count": null,
    "id": "advisory-preserve",
    "metadata": {},
    "outputs": [],
    "source": [
+    "# Create helper functions\n",
+    "\n",
     "def tile_generator(bounds: GeoBoundary, zoom : int):\n",
     "    # Returns all tiles that intersect with the provide bounds\n",
     "    bds = box(bounds.top_right.lng, bounds.top_right.lat, bounds.bottom_left.lng, bounds.bottom_left.lat).bounds\n",
@@ -296,16 +298,7 @@
     "    max_x -= tile_bounds[0]\n",
     "    min_y -= tile_bounds[1]\n",
     "    max_y -= tile_bounds[1]\n",
-    "    return [int(x) for x in [min_x, min_y, max_x, max_y]], intersection\n"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": 12,
-   "id": "according-sixth",
-   "metadata": {},
-   "outputs": [],
-   "source": [
+    "    return [int(x) for x in [min_x, min_y, max_x, max_y]], intersection\n",
     "\n",
     "def get_image(tile : Tile, url : str) -> np.ndarray:\n",
     "    # Downloads a tile\n",
@@ -320,6 +313,16 @@
     "\n"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "informative-happiness",
+   "metadata": {},
+   "source": [
+    "* <b>Visualize tile and labeling region</b>\n",
+    "* The tile is one of the tiles that overlapped with the `GeoBoundary` that we provided\n",
+    "* The red box, `GeoBoundary`, is the area where the labelers will label"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": 13,
@@ -372,6 +375,16 @@
     "Image.fromarray(crop)"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "distant-breast",
+   "metadata": {},
+   "source": [
+    "* <b>Run Model</b>\n",
+    "* This is where users can run model inferences.\n",
+    "* For this example we are detecting water with some simple image processing"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": 15,
@@ -411,6 +424,14 @@
     "Image.fromarray(mask)"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "earned-correspondence",
+   "metadata": {},
+   "source": [
+    "* <b>Convert Mask to WGS84 Polygons<b>"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": 17,
@@ -419,6 +440,7 @@
    "outputs": [],
    "source": [
     "def extract_polygons(mask: np.ndarray) -> Polygon:\n",
+    "    # Turns a segementation mask into shapely polygons\n",
     "    contours, _ = cv2.findContours(mask, cv2.RETR_TREE, cv2.CHAIN_APPROX_NONE)\n",
     "    polygons = []\n",
     "\n",
@@ -431,7 +453,8 @@
     "    polygons = [Polygon(poly) for poly in polygons]\n",
     "    return [poly.simplify(3) for poly in polygons if poly.area > 50]\n",
     "\n",
-    "def convert_to_epsg4326(poly : Polygon, intersection: Polygon) -> List[GeoPoint]:\n",
+    "def convert_to_wgs84(poly : Polygon, intersection: Polygon) -> List[GeoPoint]:\n",
+    "    # Turns pixel polygons into wgs84 polygons\n",
     "    result = []\n",
     "    for coord in poly.__geo_interface__['coordinates'][0]:\n",
     "        x = intersection.bounds[1] + coord[0]\n",
@@ -478,7 +501,7 @@
    "metadata": {},
    "outputs": [],
    "source": [
-    "geo_polygons = [convert_to_epsg4326(poly, intersection) for poly in pixel_polygons]"
+    "geo_polygons = [convert_to_wgs84(poly, intersection) for poly in pixel_polygons]"
    ]
   },
   {
@@ -508,6 +531,14 @@
     "geo_polygons"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "industrial-tuner",
+   "metadata": {},
+   "source": [
+    "* <b>Create NDJson From Geo Polygons</b>"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": 27,
@@ -532,6 +563,14 @@
     "    })"
    ]
   },
+  {
+   "cell_type": "markdown",
+   "id": "former-skiing",
+   "metadata": {},
+   "source": [
+    "* <b>Upload to Labelbox</b>"
+   ]
+  },
   {
    "cell_type": "code",
    "execution_count": 28,