Merge branch 'master' of github.com:developmentseed/label-maker

Author: drewbo

.circleci/config.yml

@@ -47,7 +47,16 @@ jobs:
             cp -r test/fixtures/integration/tiles integration-sg/tiles
             label-maker labels --dest integration-sg --config test/fixtures/integration/config.integration.segmentation.json
             python test/verify-labels-sg.py
-
+            
+            # object detection integration tests
+            mkdir integration-od
+            cp -r test/fixtures/integration/tiles integration-od/tiles
+            cp test/fixtures/integration/portugal-z17.mbtiles integration-od/
+            label-maker labels -d integration-od -c test/fixtures/integration/config.integration.object_detection.json
+            python test/verify-labels-od.py
+            label-maker package -d integration-od -c test/fixtures/integration/config.integration.object_detection.json
+            python test/verify-package-od.py
+            
             # Retest that `label-maker labels` works, but from outside the module directory
             cd ~
             label-maker labels --dest label-maker/integration-cl --config label-maker/test/fixtures/integration/config.integration.json > stdout

label_maker/package.py

@@ -6,6 +6,7 @@
 import numpy as np
 from PIL import Image
 
+
 def package_directory(dest_folder, classes, imagery, ml_type, seed=False, train_size=0.8, **kwargs):
     """Generate an .npz file containing arrays for training machine learning algorithms
 
@@ -44,8 +45,8 @@ def package_directory(dest_folder, classes, imagery, ml_type, seed=False, train_
     # find maximum number of features in advance so numpy shapes match
     if ml_type == 'object-detection':
         max_features = 0
-        for tile in tiles.files:
-            features = len(tiles[tile][0])
+        for tile in labels.files:
+            features = len(labels[tile])
             if features > max_features:
                 max_features = features
 

test/fixtures/integration/config.integration.object_detection.json

@@ -0,0 +1,22 @@
+{
+  "country": "portugal",
+  "bounding_box": [
+    -9.4575,
+    38.8467,
+    -9.4510,
+    38.8513
+  ],
+  "zoom": 17,
+  "classes": [
+    { "name": "Water Tower", "filter": ["==", "man_made", "water_tower"] },
+    { "name": "Building", "filter": ["has", "building"] },
+    { "name": "Farmland", "filter": ["==", "landuse", "farmland"] },
+    { "name": "Ruins", "filter": ["==", "historic", "ruins"] },
+    { "name": "Parking", "filter": ["==", "amenity", "parking"] },
+    { "name": "Roads", "filter": ["has", "highway"] }
+  ],
+  "imagery": "https://api.mapbox.com/v4/mapbox.satellite/{z}/{x}/{y}.jpg?access_token=ACCESS_TOKEN",
+  "background_ratio": 1,
+  "ml_type": "object-detection",
+  "seed": 19
+}

test/verify-labels-od.py

@@ -0,0 +1,50 @@
+"""Validate that the output produced by integration testing on 'label-maker labels'
+matches our expectations, specifically for object detection tasks"""
+import numpy as np
+
+labels = np.load("integration-od/labels.npz")
+
+expected_num = dict()
+expected_num['62092-50162-17'] = np.array([])
+expected_num['62092-50163-17'] = np.array([
+    [209, 192, 259, 259, 6], [251, 251, 259, 259, 6]
+])
+expected_num['62092-50164-17'] = np.array([
+    [209, -4, 250, 28, 6], [242, -4, 259, 28, 6],
+    [222, 13, 235, 66, 6], [87, 20, 250, 259, 6]
+])
+expected_num['62093-50162-17'] = np.array([
+    [81, 145, 128, 259, 6], [124, -4, 218, 259, 6],
+    [207, -4, 247, 153, 6], [140, 108, 193, 259, 6],
+    [125, 236, 152, 259, 6], [162, 177, 176, 216, 6],
+    [170, 151, 214, 179, 6], [141, 166, 244, 259, 6],
+    [203, 88, 259, 186, 6]
+])
+expected_num['62093-50163-17'] = np.array([
+    [81, -4, 125, 15, 6], [117, -4, 133, 17, 6],
+    [119, -4, 151, 36, 6], [125, -4, 140, 7, 6],
+    [141, -4, 187, 7, 6], [64, 32, 91, 60, 4],
+    [84, 50, 106, 64, 6], [111, 9, 127, 26, 6],
+    [111, 18, 127, 35, 6], [84, 15, 119, 52, 6],
+    [74, 6, 129, 69, 5], [93, 24, 123, 46, 6],
+    [88, 27, 127, 93, 6], [-4, 85, 96, 213, 6],
+    [-2, 85, 96, 259, 6], [115, 38, 259, 100, 6]
+])
+expected_num['62094-50162-17'] = np.array([
+    [67, -4, 172, 248, 6], [-4, 172, 90, 259, 6],
+    [91, 170, 259, 227, 6]
+])
+expected_num['62093-50164-17'] = np.array([
+    [-4, -4, 12, 22, 6], [207, 158, 259, 195, 6]
+])
+expected_num['62094-50163-17'] = np.array([
+    [73, -4, 259, 78, 6], [30, 166, 60, 196, 1],
+    [30, 166, 60, 196, 2], [203, 129, 259, 259, 6],
+    [-4, 90, 259, 138, 6]
+])
+expected_num['62094-50164-17'] = np. array([
+    [158, -4, 216, 82, 6], [-4, 108, 147, 173, 6],
+    [139, 74, 254, 143, 6], [240, 90, 259, 232, 6]
+])
+
+assert len(labels.files) == len(expected_num.keys())

test/verify-package-od.py

@@ -0,0 +1,13 @@
+"""Validate that the output produced by integration testing on 'label-maker package' matches our expectations"""
+import numpy as np
+
+data = np.load('integration-od/data.npz')
+
+assert np.sum(data['x_train']) == 144752757
+assert np.sum(data['x_test']) == 52758414
+assert data['x_train'].shape == (6, 256, 256, 3)
+assert data['x_test'].shape == (2, 256, 256, 3)
+
+# validate our label data with exact matches in shape
+assert data['y_train'].shape == (6, 16, 5)
+assert data['y_test'].shape ==  (2, 16, 5)