Merge pull request #673 from Labelbox/jtso/al-3329

[AL-3329] Document Annotation Type for Export

Author: jtsodapop

labelbox/data/serialization/labelbox_v1/label.py

@@ -146,6 +146,7 @@ class LBV1Label(BaseModel):
     skipped: Optional[bool] = Extra('Skipped')
     media_type: Optional[str] = Extra('media_type')
     data_split: Optional[str] = Extra('Data Split')
+    global_key: Optional[str] = Extra('Global Key')
 
     def to_common(self) -> Label:
         if isinstance(self.label, list):

labelbox/data/serialization/labelbox_v1/objects.py

@@ -22,6 +22,8 @@ class LBV1ObjectBase(LBV1Feature):
     instanceURI: Optional[str] = None
     classifications: List[Union[LBV1Text, LBV1Radio, LBV1Dropdown,
                                 LBV1Checklist]] = []
+    page: Optional[int] = None
+    unit: Optional[str] = None
 
     def dict(self, *args, **kwargs) -> Dict[str, Any]:
         res = super().dict(*args, **kwargs)
@@ -262,7 +264,7 @@ def from_common(cls, text_entity: TextEntity,
 class LBV1Objects(BaseModel):
     objects: List[Union[LBV1Line, LBV1Point, LBV1Polygon, LBV1Rectangle,
                         LBV1TextEntity, LBV1Mask, LBV1TIPoint, LBV1TILine,
-                        LBV1TIPolygon, LBV1TIRectangle]]
+                        LBV1TIPolygon, LBV1TIRectangle,]]
 
     def to_common(self) -> List[ObjectAnnotation]:
         objects = [
@@ -285,6 +287,8 @@ def to_common(self) -> List[ObjectAnnotation]:
                                  'color': obj.color,
                                  'feature_id': obj.feature_id,
                                  'value': obj.value,
+                                 'page': obj.page,
+                                 'unit': obj.unit,
                              }) for obj in self.objects
         ]
         return objects

tests/data/assets/labelbox_v1/pdf_export.json

@@ -0,0 +1,130 @@
+[{
+    "ID": "cl6xnzi4a7ldn0729381g7104",
+    "DataRow ID": "cl6xnv9h61fv0085yhtoq06ht",
+    "Labeled Data": "https://storage.labelbox.com/ckcz6bubudyfi0855o1dt1g9s%2F4cef4e08-e13d-8a5e-fbbf-c7624babb490-Airbnb_%20Labelbox%20-%20Focus%20on%20Workforce%20-%20Labelbox%20Labeling%20Operations%20(1).pdf?Expires=1661971050348&KeyName=labelbox-assets-key-3&Signature=JK6ral5CXF7T9Q5LaQqKvJy5A2A",
+    "Label": {
+        "objects": [{
+            "featureId": "cl6xnzjpq0dmr07yocs2vfot8",
+            "schemaId": "cl6xnuwt95lqq07330tbb3mfd",
+            "color": "#1CE6FF",
+            "title": "boxy",
+            "value": "boxy",
+            "bbox": {
+                "top": 144.68,
+                "left": 107.84,
+                "height": 441.6,
+                "width": 9.48
+            },
+            "page": 0,
+            "unit": "POINTS",
+            "instanceURI": "https://api.labelbox.com/masks/feature/cl6xnzjpq0dmr07yocs2vfot8?token=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJ1c2VySWQiOiJja2NjOWZtbXc0aGNkMDczOHFpeWM2YW54Iiwib3JnYW5pemF0aW9uSWQiOiJja2N6NmJ1YnVkeWZpMDg1NW8xZHQxZzlzIiwiaWF0IjoxNjYwNzYxNDUwLCJleHAiOjE2NjMzNTM0NTB9.X4-j6zee8o685PUrL9C6oC2m6TayKuJQHhN8iLgG8kI"
+        }, {
+            "featureId": "cl6xnzjpq0dms07yobwv68gxf",
+            "schemaId": "cl6xnuwt95lqq07330tbb3mfd",
+            "color": "#1CE6FF",
+            "title": "boxy",
+            "value": "boxy",
+            "bbox": {
+                "top": 162.73,
+                "left": 32.45,
+                "height": 388.17,
+                "width": 101.66
+            },
+            "page": 4,
+            "unit": "POINTS",
+            "instanceURI": "https://api.labelbox.com/masks/feature/cl6xnzjpq0dms07yobwv68gxf?token=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJ1c2VySWQiOiJja2NjOWZtbXc0aGNkMDczOHFpeWM2YW54Iiwib3JnYW5pemF0aW9uSWQiOiJja2N6NmJ1YnVkeWZpMDg1NW8xZHQxZzlzIiwiaWF0IjoxNjYwNzYxNDUwLCJleHAiOjE2NjMzNTM0NTB9.X4-j6zee8o685PUrL9C6oC2m6TayKuJQHhN8iLgG8kI"
+        }, {
+            "featureId": "cl6xnzjpq0dmt07yo8pp45gru",
+            "schemaId": "cl6xnuwt95lqq07330tbb3mfd",
+            "color": "#1CE6FF",
+            "title": "boxy",
+            "value": "boxy",
+            "bbox": {
+                "top": 223.26,
+                "left": 251.42,
+                "height": 457.04,
+                "width": 186.78
+            },
+            "page": 7,
+            "unit": "POINTS",
+            "instanceURI": "https://api.labelbox.com/masks/feature/cl6xnzjpq0dmt07yo8pp45gru?token=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJ1c2VySWQiOiJja2NjOWZtbXc0aGNkMDczOHFpeWM2YW54Iiwib3JnYW5pemF0aW9uSWQiOiJja2N6NmJ1YnVkeWZpMDg1NW8xZHQxZzlzIiwiaWF0IjoxNjYwNzYxNDUwLCJleHAiOjE2NjMzNTM0NTB9.X4-j6zee8o685PUrL9C6oC2m6TayKuJQHhN8iLgG8kI"
+        }, {
+            "featureId": "cl6xnzjpq0dmu07yo2qik0en4",
+            "schemaId": "cl6xnuwt95lqq07330tbb3mfd",
+            "color": "#1CE6FF",
+            "title": "boxy",
+            "value": "boxy",
+            "bbox": {
+                "top": 32.52,
+                "left": 218.17,
+                "height": 231.73,
+                "width": 110.56
+            },
+            "page": 6,
+            "unit": "POINTS",
+            "instanceURI": "https://api.labelbox.com/masks/feature/cl6xnzjpq0dmu07yo2qik0en4?token=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJ1c2VySWQiOiJja2NjOWZtbXc0aGNkMDczOHFpeWM2YW54Iiwib3JnYW5pemF0aW9uSWQiOiJja2N6NmJ1YnVkeWZpMDg1NW8xZHQxZzlzIiwiaWF0IjoxNjYwNzYxNDUwLCJleHAiOjE2NjMzNTM0NTB9.X4-j6zee8o685PUrL9C6oC2m6TayKuJQHhN8iLgG8kI"
+        }, {
+            "featureId": "cl6xnzjpq0dmv07yo7phz7ofz",
+            "schemaId": "cl6xnuwt95lqq07330tbb3mfd",
+            "color": "#1CE6FF",
+            "title": "boxy",
+            "value": "boxy",
+            "bbox": {
+                "top": 117.39,
+                "left": 4.25,
+                "height": 456.92,
+                "width": 164.83
+            },
+            "page": 7,
+            "unit": "POINTS",
+            "instanceURI": "https://api.labelbox.com/masks/feature/cl6xnzjpq0dmv07yo7phz7ofz?token=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJ1c2VySWQiOiJja2NjOWZtbXc0aGNkMDczOHFpeWM2YW54Iiwib3JnYW5pemF0aW9uSWQiOiJja2N6NmJ1YnVkeWZpMDg1NW8xZHQxZzlzIiwiaWF0IjoxNjYwNzYxNDUwLCJleHAiOjE2NjMzNTM0NTB9.X4-j6zee8o685PUrL9C6oC2m6TayKuJQHhN8iLgG8kI"
+        }, {
+            "featureId": "cl6xnzjpq0dmw07yofocp6uf6",
+            "schemaId": "cl6xnuwt95lqq07330tbb3mfd",
+            "color": "#1CE6FF",
+            "title": "boxy",
+            "value": "boxy",
+            "bbox": {
+                "top": 82.13,
+                "left": 217.28,
+                "height": 279.76,
+                "width": 82.43
+            },
+            "page": 8,
+            "unit": "POINTS",
+            "instanceURI": "https://api.labelbox.com/masks/feature/cl6xnzjpq0dmw07yofocp6uf6?token=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJ1c2VySWQiOiJja2NjOWZtbXc0aGNkMDczOHFpeWM2YW54Iiwib3JnYW5pemF0aW9uSWQiOiJja2N6NmJ1YnVkeWZpMDg1NW8xZHQxZzlzIiwiaWF0IjoxNjYwNzYxNDUwLCJleHAiOjE2NjMzNTM0NTB9.X4-j6zee8o685PUrL9C6oC2m6TayKuJQHhN8iLgG8kI"
+        }, {
+            "featureId": "cl6xnzjpq0dmx07yo0qh40z0n",
+            "schemaId": "cl6xnuwt95lqq07330tbb3mfd",
+            "color": "#1CE6FF",
+            "title": "boxy",
+            "value": "boxy",
+            "bbox": {
+                "top": 298.12,
+                "left": 83.34,
+                "height": 203.83,
+                "width": 0.38
+            },
+            "page": 3,
+            "unit": "POINTS",
+            "instanceURI": "https://api.labelbox.com/masks/feature/cl6xnzjpq0dmx07yo0qh40z0n?token=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJ1c2VySWQiOiJja2NjOWZtbXc0aGNkMDczOHFpeWM2YW54Iiwib3JnYW5pemF0aW9uSWQiOiJja2N6NmJ1YnVkeWZpMDg1NW8xZHQxZzlzIiwiaWF0IjoxNjYwNzYxNDUwLCJleHAiOjE2NjMzNTM0NTB9.X4-j6zee8o685PUrL9C6oC2m6TayKuJQHhN8iLgG8kI"
+        }],
+        "classifications": [],
+        "relationships": []
+    },
+    "Created By": "jtso@labelbox.com",
+    "Project Name": "PDF MAL Test",
+    "Created At": "2022-08-17T18:37:18.000Z",
+    "Updated At": "2022-08-17T18:37:20.073Z",
+    "Seconds to Label": 15.003,
+    "External ID": "Airbnb_ Labelbox - Focus on Workforce - Labelbox Labeling Operations (1).pdf",
+    "Global Key": null,
+    "Agreement": -1,
+    "Benchmark Agreement": -1,
+    "Benchmark ID": null,
+    "Dataset Name": "PDF ",
+    "Reviews": [],
+    "View Label": "https://editor.labelbox.com?project=cl6xntneb7t28072bggdydv7a&label=cl6xnzi4a7ldn0729381g7104",
+    "Has Open Issues": 0,
+    "Skipped": false
+}]

tests/data/serialization/labelbox_v1/test_document.py

@@ -0,0 +1,45 @@
+import json
+from typing import Dict, Any
+
+from labelbox.data.serialization.labelbox_v1.converter import LBV1Converter
+
+IGNORE_KEYS = [
+    "Data Split", "media_type", "DataRow Metadata", "Media Attributes"
+]
+
+
+def round_dict(data: Dict[str, Any]) -> Dict[str, Any]:
+    for key in data:
+        if isinstance(data[key], float):
+            data[key] = int(data[key])
+        elif isinstance(data[key], dict):
+            data[key] = round_dict(data[key])
+    return data
+
+
+def test_pdf():
+    """
+    Tests an export from a pdf document with only bounding boxes
+    """
+    payload = json.load(
+        open('tests/data/assets/labelbox_v1/pdf_export.json', 'r'))
+    collection = LBV1Converter.deserialize(payload)
+    serialized = next(LBV1Converter.serialize(collection))
+
+    payload = payload[0]  # only one document in the export
+
+    serialized = {k: v for k, v in serialized.items() if k not in IGNORE_KEYS}
+
+    assert serialized.keys() == payload.keys()
+    for key in payload.keys():
+        if key == 'Label':
+            serialized_no_classes = [{
+                k: v for k, v in dic.items() if k != 'classifications'
+            } for dic in serialized[key]['objects']]
+            serialized_round = [
+                round_dict(dic) for dic in serialized_no_classes
+            ]
+            payload_round = [round_dict(dic) for dic in payload[key]['objects']]
+            assert payload_round == serialized_round
+        else:
+            assert serialized[key] == payload[key]

tests/data/serialization/labelbox_v1/test_image.py

@@ -22,6 +22,7 @@ def test_image(file_path):
 
     # We are storing the media types now.
     payload['media_type'] = 'image'
+    payload['Global Key'] = None
 
     assert serialized.keys() == payload.keys()
 
@@ -31,6 +32,8 @@ def test_image(file_path):
         elif key == 'Label':
             for annotation_a, annotation_b in zip(serialized[key]['objects'],
                                                   payload[key]['objects']):
+                annotation_b['page'] = None
+                annotation_b['unit'] = None
                 if not len(annotation_a['classifications']):
                     # We don't add a classification key to the payload if there is no classifications.
                     annotation_a.pop('classifications')

tests/data/serialization/labelbox_v1/test_text.py

@@ -10,6 +10,7 @@ def test_text():
     serialized = next(LBV1Converter.serialize(collection))
 
     payload['media_type'] = 'text'
+    payload['Global Key'] = None
 
     assert serialized.keys() == payload.keys()
     for key in serialized:
@@ -18,6 +19,8 @@ def test_text():
         elif key == 'Label':
             for annotation_a, annotation_b in zip(serialized[key]['objects'],
                                                   payload[key]['objects']):
+                annotation_b['page'] = None
+                annotation_b['unit'] = None
                 if not len(annotation_a['classifications']):
                     # We don't add a classification key to the payload if there is no classifications.
                     annotation_a.pop('classifications')

tests/data/serialization/labelbox_v1/test_unknown_media.py

@@ -16,6 +16,7 @@ def test_image():
 
     for row in payload:
         row['media_type'] = 'image'
+        row['Global Key'] = None
 
     collection = LBV1Converter.deserialize(payload)
     for idx, serialized in enumerate(LBV1Converter.serialize(collection)):
@@ -30,6 +31,8 @@ def test_image():
                     if not len(annotation_a['classifications']):
                         # We don't add a classification key to the payload if there is no classifications.
                         annotation_a.pop('classifications')
+                    annotation_b['page'] = None
+                    annotation_b['unit'] = None
 
                     if isinstance(annotation_b.get('classifications'),
                                   list) and len(

tests/data/serialization/labelbox_v1/test_video.py

@@ -18,6 +18,7 @@ def test_video():
     collection = LBV1Converter.deserialize([payload])
     serialized = next(LBV1Converter.serialize(collection))
     payload['media_type'] = 'video'
+    payload['Global Key'] = None
     assert serialized.keys() == payload.keys()
     for key in serialized:
         if key != 'Label':
@@ -32,6 +33,8 @@ def test_video():
 
                 for obj_a, obj_b in zip(annotation_a['objects'],
                                         annotation_b['objects']):
+                    obj_b['page'] = None
+                    obj_b['unit'] = None
                     obj_a = round_dict(obj_a)
                     obj_b = round_dict(obj_b)
                     assert obj_a == obj_b