Global keys to Annotation Types, and add validation when both id and globalKey are provided

Author: Kevin Kim

labelbox/data/annotation_types/data/raster.py

@@ -163,9 +163,10 @@ def validate_args(cls, values):
         url = values.get("url")
         arr = values.get("arr")
         uid = values.get('uid')
-        if uid == file_path == im_bytes == url == None and arr is None:
+        global_key = values.get('global_key')
+        if uid == file_path == im_bytes == url == global_key == None and arr is None:
             raise ValueError(
-                "One of `file_path`, `im_bytes`, `url`, `uid` or `arr` required."
+                "One of `file_path`, `im_bytes`, `url`, `uid`, `global_key` or `arr` required."
             )
         if arr is not None:
             if arr.dtype != np.uint8:

labelbox/data/annotation_types/data/text.py

@@ -93,9 +93,11 @@ def validate_date(cls, values):
         text = values.get("text")
         url = values.get("url")
         uid = values.get('uid')
-        if uid == file_path == text == url == None:
+        global_key = values.get('global_key')
+        if uid == file_path == text == url == global_key == None:
             raise ValueError(
-                "One of `file_path`, `text`, `uid`, or `url` required.")
+                "One of `file_path`, `text`, `uid`, `global_key` or `url` required."
+            )
         return values
 
     def __repr__(self) -> str:

labelbox/data/annotation_types/data/video.py

@@ -153,10 +153,12 @@ def validate_data(cls, values):
         url = values.get("url")
         frames = values.get("frames")
         uid = values.get("uid")
+        global_key = values.get("global_key")
 
-        if uid == file_path == frames == url == None:
+        if uid == file_path == frames == url == global_key == None:
             raise ValueError(
-                "One of `file_path`, `frames`, `uid`, or `url` required.")
+                "One of `file_path`, `frames`, `uid`, `global_key` or `url` required."
+            )
         return values
 
     def __repr__(self) -> str:

labelbox/data/serialization/ndjson/base.py

@@ -8,15 +8,17 @@
 
 class DataRow(BaseModel):
     id: str = None
+    global_key: str = None
 
-    @validator('id', pre=True, always=True)
-    def validate_id(cls, v):
-        if v is None:
-            raise ValueError(
-                "Data row ids are not set. Use `LabelGenerator.add_to_dataset`,or `Label.create_data_row`. "
-                "You can also manually assign the id for each `BaseData` object"
-            )
-        return v
+    @root_validator()
+    def must_set_one(cls, values):
+        if bool(values.get('id')) == bool(values.get('global_key')):
+            raise ValueError("Must set either id or global_key")
+        return values
+
+    class Config:
+        allow_population_by_field_name = True
+        alias_generator = camel_case
 
 
 class NDJsonBase(BaseModel):
@@ -27,6 +29,15 @@ class NDJsonBase(BaseModel):
     def set_id(cls, v):
         return v or str(uuid4())
 
+    def dict(self, *args, **kwargs):
+        """ Pop missing id or missing globalKey from dataRow """
+        res = super().dict(*args, **kwargs)
+        if not self.data_row.id:
+            res['dataRow'].pop('id')
+        if not self.data_row.global_key:
+            res['dataRow'].pop('globalKey')
+        return res
+
     class Config:
         allow_population_by_field_name = True
         alias_generator = camel_case

labelbox/data/serialization/ndjson/classification.py

@@ -8,7 +8,7 @@
 from ...annotation_types.classification.classification import ClassificationAnswer, Dropdown, Text, Checklist, Radio
 from ...annotation_types.types import Cuid
 from ...annotation_types.data import TextData, VideoData, ImageData
-from .base import NDAnnotation
+from .base import DataRow, NDAnnotation
 
 
 class NDFeature(ConfidenceMixin):
@@ -125,7 +125,7 @@ def from_common(cls, text: Text, name: str, feature_schema_id: Cuid,
                                                        ImageData]) -> "NDText":
         return cls(
             answer=text.answer,
-            data_row={'id': data.uid},
+            data_row=DataRow(id=data.uid, global_key=data.global_key),
             name=name,
             schema_id=feature_schema_id,
             uuid=extra.get('uuid'),
@@ -145,7 +145,7 @@ def from_common(
                       confidence=answer.confidence)
             for answer in checklist.answer
         ],
-                   data_row={'id': data.uid},
+                   data_row=DataRow(id=data.uid, global_key=data.global_key),
                    name=name,
                    schema_id=feature_schema_id,
                    uuid=extra.get('uuid'),
@@ -161,7 +161,7 @@ def from_common(cls, radio: Radio, name: str, feature_schema_id: Cuid,
         return cls(answer=NDFeature(name=radio.answer.name,
                                     schema_id=radio.answer.feature_schema_id,
                                     confidence=radio.answer.confidence),
-                   data_row={'id': data.uid},
+                   data_row=DataRow(id=data.uid, global_key=data.global_key),
                    name=name,
                    schema_id=feature_schema_id,
                    uuid=extra.get('uuid'),

labelbox/data/serialization/ndjson/label.py

@@ -1,6 +1,6 @@
 from itertools import groupby
 from operator import itemgetter
-from typing import Dict, Generator, List, Tuple, Union
+from typing import Dict, Generator, List, Optional, Tuple, Union
 from collections import defaultdict
 import warnings
 
@@ -17,6 +17,7 @@
 from .metric import NDScalarMetric, NDMetricAnnotation, NDConfusionMatrixMetric
 from .classification import NDChecklistSubclass, NDClassification, NDClassificationType, NDRadioSubclass
 from .objects import NDObject, NDObjectType, NDSegments
+from .base import DataRow
 
 
 class NDLabel(BaseModel):
@@ -27,7 +28,10 @@ class NDLabel(BaseModel):
     def to_common(self) -> LabelGenerator:
         grouped_annotations = defaultdict(list)
         for annotation in self.annotations:
-            grouped_annotations[annotation.data_row.id].append(annotation)
+            grouped_annotations[annotation.data_row.id or
+                                annotation.data_row.global_key].append(
+                                    annotation)
+        print(grouped_annotations)
         return LabelGenerator(
             data=self._generate_annotations(grouped_annotations))
 
@@ -45,9 +49,11 @@ def _generate_annotations(
                                              NDConfusionMatrixMetric,
                                              NDScalarMetric, NDSegments]]]
     ) -> Generator[Label, None, None]:
-        for data_row_id, annotations in grouped_annotations.items():
+        for _, annotations in grouped_annotations.items():
             annots = []
+            data_row = annotations[0].data_row
             for annotation in annotations:
+
                 if isinstance(annotation, NDSegments):
                     annots.extend(
                         NDSegments.to_common(annotation, annotation.name,
@@ -62,22 +68,30 @@ def _generate_annotations(
                 else:
                     raise TypeError(
                         f"Unsupported annotation. {type(annotation)}")
-            data = self._infer_media_type(annots)(uid=data_row_id)
-            yield Label(annotations=annots, data=data)
+            yield Label(annotations=annots,
+                        data=self._infer_media_type(data_row, annots))
 
     def _infer_media_type(
-        self, annotations: List[Union[TextEntity, VideoClassificationAnnotation,
-                                      VideoObjectAnnotation, ObjectAnnotation,
-                                      ClassificationAnnotation, ScalarMetric,
-                                      ConfusionMatrixMetric]]
+        self, data_row: DataRow,
+        annotations: List[Union[TextEntity, VideoClassificationAnnotation,
+                                VideoObjectAnnotation, ObjectAnnotation,
+                                ClassificationAnnotation, ScalarMetric,
+                                ConfusionMatrixMetric]]
     ) -> Union[TextData, VideoData, ImageData]:
+        if len(annotations) == 0:
+            raise ValueError("Missing annotations while inferring media type")
+
         types = {type(annotation) for annotation in annotations}
+        data = ImageData
         if TextEntity in types:
-            return TextData
+            data = TextData
         elif VideoClassificationAnnotation in types or VideoObjectAnnotation in types:
-            return VideoData
+            data = VideoData
+
+        if data_row.id:
+            return data(uid=data_row.id)
         else:
-            return ImageData
+            return data(global_key=data_row.global_key)
 
     @staticmethod
     def _get_consecutive_frames(

labelbox/data/serialization/ndjson/metric.py

@@ -1,7 +1,7 @@
 from typing import Optional, Union, Type
 
 from labelbox.data.annotation_types.data import ImageData, TextData
-from labelbox.data.serialization.ndjson.base import NDJsonBase
+from labelbox.data.serialization.ndjson.base import DataRow, NDJsonBase
 from labelbox.data.annotation_types.metrics.scalar import (
     ScalarMetric, ScalarMetricAggregation, ScalarMetricValue,
     ScalarMetricConfidenceValue)
@@ -50,7 +50,7 @@ def from_common(
                    feature_name=metric.feature_name,
                    subclass_name=metric.subclass_name,
                    aggregation=metric.aggregation,
-                   data_row={'id': data.uid})
+                   data_row=DataRow(id=data.uid, global_key=data.global_key))
 
 
 class NDScalarMetric(BaseNDMetric):
@@ -75,7 +75,7 @@ def from_common(cls, metric: ScalarMetric,
                    feature_name=metric.feature_name,
                    subclass_name=metric.subclass_name,
                    aggregation=metric.aggregation.value,
-                   data_row={'id': data.uid})
+                   data_row=DataRow(id=data.uid, global_key=data.global_key))
 
     def dict(self, *args, **kwargs):
         res = super().dict(*args, **kwargs)

labelbox/data/serialization/ndjson/objects.py

@@ -60,7 +60,7 @@ def from_common(cls,
             'x': point.x,
             'y': point.y
         },
-                   dataRow=DataRow(id=data.uid),
+                   data_row=DataRow(id=data.uid, global_key=data.global_key),
                    name=name,
                    schema_id=feature_schema_id,
                    uuid=extra.get('uuid'),
@@ -105,7 +105,7 @@ def from_common(cls,
             'x': pt.x,
             'y': pt.y
         } for pt in line.points],
-                   dataRow=DataRow(id=data.uid),
+                   data_row=DataRow(id=data.uid, global_key=data.global_key),
                    name=name,
                    schema_id=feature_schema_id,
                    uuid=extra.get('uuid'),
@@ -154,7 +154,7 @@ def from_common(cls,
             'x': pt.x,
             'y': pt.y
         } for pt in polygon.points],
-                   dataRow=DataRow(id=data.uid),
+                   data_row=DataRow(id=data.uid, global_key=data.global_key),
                    name=name,
                    schema_id=feature_schema_id,
                    uuid=extra.get('uuid'),
@@ -183,7 +183,7 @@ def from_common(cls,
                              left=rectangle.start.x,
                              height=rectangle.end.y - rectangle.start.y,
                              width=rectangle.end.x - rectangle.start.x),
-                   dataRow=DataRow(id=data.uid),
+                   data_row=DataRow(id=data.uid, global_key=data.global_key),
                    name=name,
                    schema_id=feature_schema_id,
                    uuid=extra.get('uuid'),
@@ -280,7 +280,7 @@ def from_common(cls, segments: List[VideoObjectAnnotation], data: VideoData,
         segments = [NDSegment.from_common(segment) for segment in segments]
 
         return cls(segments=segments,
-                   dataRow=DataRow(id=data.uid),
+                   data_row=DataRow(id=data.uid, global_key=data.global_key),
                    name=name,
                    schema_id=feature_schema_id,
                    uuid=extra.get('uuid'))
@@ -332,7 +332,7 @@ def from_common(cls,
                 png=base64.b64encode(im_bytes.getvalue()).decode('utf-8'))
 
         return cls(mask=lbv1_mask,
-                   dataRow=DataRow(id=data.uid),
+                   data_row=DataRow(id=data.uid, global_key=data.global_key),
                    name=name,
                    schema_id=feature_schema_id,
                    uuid=extra.get('uuid'),
@@ -364,7 +364,7 @@ def from_common(cls,
             start=text_entity.start,
             end=text_entity.end,
         ),
-                   dataRow=DataRow(id=data.uid),
+                   data_row=DataRow(id=data.uid, global_key=data.global_key),
                    name=name,
                    schema_id=feature_schema_id,
                    uuid=extra.get('uuid'),

labelbox/schema/annotation_import.py

@@ -155,7 +155,7 @@ def _get_ndjson_from_objects(cls, objects: Union[List[Dict[str, Any]],
             )
 
         objects = serialize_labels(objects)
-        # cls.validate_data_rows(objects)
+        cls.validate_data_rows(objects)
 
         data_str = ndjson.dumps(objects)
         if not data_str: