Addressing comments from pull request, basic clean-up, added info abo…

…ut UTM reprojections

cerulean_cloud/cloud_run_orchestrator/handler.py

@@ -340,7 +340,7 @@ async def _orchestrate(
     print(f"Offset 2 offset_group_bounds: {offset_2_group_bounds}")
 
     print(
-        f"Original tiles are {len(base_tiles)}, {len(offset_group_bounds)}, {len(offset_2_group_bounds)}"
+        f"Original tiles are {len(base_tiles)}, {len(offset_tiles_bounds)}, {len(offset_2_tiles_bounds)}"
     )
 
     # Filter out land tiles
@@ -425,7 +425,7 @@ async def _orchestrate(
 
             offset_2_tiles_inference = await perform_inference(
                 offset_2_tiles_bounds,
-                cloud_run_inference.get_offset_tile_inference,  # THIS FUNCTION NEEDS TO BE EDITTED
+                cloud_run_inference.get_offset_tile_inference,
                 20,
                 "offset2 tiles",
             )
@@ -551,7 +551,7 @@ async def _orchestrate(
             async with db_client.session.begin():
                 orchestrator_run.success = True
                 orchestrator_run.inference_end_time = end_time
-            # XXXC >> reduce num variables to a list of featureCollection, ala the merging approach
+
             orchestrator_result = OrchestratorResult(
                 classification_base=out_fc,
                 classification_offset=out_fc_offset,

cerulean_cloud/cloud_run_orchestrator/merging.py

@@ -28,14 +28,16 @@ def merge_inferences(
 
     Parameters:
     - feature_collections: A list of FeatureCollecitons to be merged, a primary and any secondary FeatureCollections
-    - isolated_conf_multiplier: A multiplier for the confidence of isolated features (default is 1 / len(feature_collections)).
     - proximity_meters: The distance to check for neighboring features and expand the geometries (default is 500m).
     - closing_meters: The distance to apply the morphological 'closing' operation (default is 0m).
     - opening_meters: The distance to apply the morphological 'opening' operation (default is 0m).
 
     Returns:
     A merged geojson FeatureCollection.
     """
+    # We reproject to UTM for processing. This assumes that all offset images will either be in the same UTM zone as
+    # the input image chip, or that the difference that arise from an offset crossing into a second UTM zone will
+    # have little or no impact on comparison to the origina image.
     gdfs_for_processing = [
         reproject_to_utm(
             gpd.GeoDataFrame.from_features(fc["features"], crs=4326).assign(fc_index=i)