Merge pull request #347 from Labelbox/develop

3.10.0

Author: msokoloff1

.gitignore

@@ -83,6 +83,9 @@ ipython_config.py
 # pyenv
 .python-version
 
+# vscode
+.vscode
+
 # pipenv
 #   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
 #   However, in case of collaboration, if having platform-specific dependencies or dependencies

CHANGELOG.md

@@ -3,10 +3,25 @@
 ## Deprecation Notice
 | Name                                  | Replacement                            | Removed After |
 | ------------------------------------- | -------------------------------------  | ------------- |
-| `ModelRun.delete_annotation_groups()` | `ModelRun.delete_model_run_data_rows()`| 3.9           |
-| `ModelRun.annotation_groups()`        | `ModelRun.model_run_data_rows()`       | 3.9           |
-| `DataRowMetadataSchema.id`            | `DataRowMetadataSchema.uid`            | 3.9           |
+| `ModelRun.delete_annotation_groups()` | `ModelRun.delete_model_run_data_rows()`|  2021-12-06   |
+| `ModelRun.annotation_groups()`        | `ModelRun.model_run_data_rows()`       |  2021-12-06   |
+| `DataRowMetadataSchema.id`            | `DataRowMetadataSchema.uid`            |  2021-12-06   |
 -----
+
+# Version 3.10.0 (2021-11-18)
+## Added
+* `AnnotationImport.wait_until_done()` accepts a `show_progress` param. This is set to `False` by default.
+    * If enabled, a tqdm progress bar will indicate the import progress.
+    * This works for all classes that inherit from AnnotationImport: `LabelImport`, `MALPredictionImport`, `MEAPredictionImport`
+    * This is not support for `BulkImportRequest` (which will eventually be replaced by `MALPredictionImport`)
+* `Option.label` and `Option.value` can now be set independently
+* `ClassificationAnswer`s now support a new `keyframe` field for videos
+* New `LBV1Label.media_type field. This is a placeholder for future backend changes.
+
+## Fix
+* Nested checklists can have extra brackets. This would cause the annotation type converter to break.
+
+
 # Version 3.9.0 (2021-11-12)
 ## Added
 * New ontology management features
@@ -18,8 +33,8 @@
     * Set up a project from an existing ontology with `project.setup_edior()`
     * Added new `FeatureSchema` entity
 * Add support for new queue modes
-    * Send batches of data direction to a project with `project.queue()`
-    * Remove items from the queue with `project.dequeue()`
+    * Send batches of data directly to a project queue with `project.queue()`
+    * Remove items from a project queue with `project.dequeue()`
     * Query for and toggle the queue mode
 
 # Version 3.8.0 (2021-10-22)

README.md

@@ -57,9 +57,9 @@ The package `rasterio` installed by `labelbox[data]` relies on GDAL which could
 You may see the following error message:
 
 ```
-INFO:root:Building on Windows requires extra options to setup.py to locate needed GDAL files. More information is available in the README. 
+INFO:root:Building on Windows requires extra options to setup.py to locate needed GDAL files. More information is available in the README.
 
-ERROR: A GDAL API version must be specified. Provide a path to gdal-config using a GDAL_CONFIG environment variable or use a GDAL_VERSION environment variable. 
+ERROR: A GDAL API version must be specified. Provide a path to gdal-config using a GDAL_CONFIG environment variable or use a GDAL_VERSION environment variable.
 ```
 
 As a workaround:
@@ -72,7 +72,7 @@ As a workaround:
 
     Note: You need to download the right files for your Python version. In the files above `cp38` means CPython 3.8.
 
-2. After downloading the files, please run the following commands, in this particular order.  
+2. After downloading the files, please run the following commands, in this particular order.
 
 ```
 pip install GDAL‑3.3.2‑cp38‑cp38‑win_amd64.wh

labelbox/__init__.py

@@ -1,5 +1,5 @@
 name = "labelbox"
-__version__ = "3.9.0"
+__version__ = "3.10.0"
 
 from labelbox.schema.project import Project
 from labelbox.client import Client

labelbox/data/annotation_types/classification/classification.py

@@ -1,4 +1,4 @@
-from typing import Any, Dict, List
+from typing import Any, Dict, List, Union, Optional
 
 try:
     from typing import Literal
@@ -24,13 +24,25 @@ class ClassificationAnswer(FeatureSchema):
     - Represents a classification option.
     - Because it inherits from FeatureSchema
         the option can be represented with either the name or feature_schema_id
+
+    - The keyframe arg only applies to video classifications.
+      Each answer can have a keyframe independent of the others.
+        So unlike object annotations, classification annotations
+          track keyframes at a classification answer level.
     """
     extra: Dict[str, Any] = {}
+    keyframe: Optional[bool] = None
+
+    def dict(self, *args, **kwargs):
+        res = super().dict(*args, **kwargs)
+        if res['keyframe'] is None:
+            res.pop('keyframe')
+        return res
 
 
 class Radio(BaseModel):
     """ A classification with only one selected option allowed
-    
+
     >>> Radio(answer = ClassificationAnswer(name = "dog"))
 
     """
@@ -50,7 +62,7 @@ class Checklist(_TempName):
 class Text(BaseModel):
     """ Free form text
 
-    >>> Text(answer = "some text answer") 
+    >>> Text(answer = "some text answer")
 
     """
     answer: str

labelbox/data/serialization/labelbox_v1/classification.py

@@ -9,56 +9,53 @@
 
 
 class LBV1ClassificationAnswer(LBV1Feature):
-    ...
+
+    def to_common(self) -> ClassificationAnswer:
+        return ClassificationAnswer(feature_schema_id=self.schema_id,
+                                    name=self.title,
+                                    keyframe=self.keyframe,
+                                    extra={
+                                        'feature_id': self.feature_id,
+                                        'value': self.value
+                                    })
+
+    @classmethod
+    def from_common(
+            cls,
+            answer: ClassificationAnnotation) -> "LBV1ClassificationAnswer":
+        return cls(schema_id=answer.feature_schema_id,
+                   title=answer.name,
+                   value=answer.extra.get('value'),
+                   feature_id=answer.extra.get('feature_id'),
+                   keyframe=answer.keyframe)
 
 
 class LBV1Radio(LBV1Feature):
     answer: LBV1ClassificationAnswer
 
     def to_common(self) -> Radio:
-        return Radio(answer=ClassificationAnswer(
-            feature_schema_id=self.answer.schema_id,
-            name=self.answer.title,
-            extra={
-                'feature_id': self.answer.feature_id,
-                'value': self.answer.value
-            }))
+        return Radio(answer=self.answer.to_common())
 
     @classmethod
     def from_common(cls, radio: Radio, feature_schema_id: Cuid,
                     **extra) -> "LBV1Radio":
         return cls(schema_id=feature_schema_id,
-                   answer=LBV1ClassificationAnswer(
-                       schema_id=radio.answer.feature_schema_id,
-                       title=radio.answer.name,
-                       value=radio.answer.extra.get('value'),
-                       feature_id=radio.answer.extra.get('feature_id')),
+                   answer=LBV1ClassificationAnswer.from_common(radio.answer),
                    **extra)
 
 
 class LBV1Checklist(LBV1Feature):
     answers: List[LBV1ClassificationAnswer]
 
     def to_common(self) -> Checklist:
-        return Checklist(answer=[
-            ClassificationAnswer(feature_schema_id=answer.schema_id,
-                                 name=answer.title,
-                                 extra={
-                                     'feature_id': answer.feature_id,
-                                     'value': answer.value
-                                 }) for answer in self.answers
-        ])
+        return Checklist(answer=[answer.to_common() for answer in self.answers])
 
     @classmethod
     def from_common(cls, checklist: Checklist, feature_schema_id: Cuid,
                     **extra) -> "LBV1Checklist":
         return cls(schema_id=feature_schema_id,
                    answers=[
-                       LBV1ClassificationAnswer(
-                           schema_id=answer.feature_schema_id,
-                           title=answer.name,
-                           value=answer.extra.get('value'),
-                           feature_id=answer.extra.get('feature_id'))
+                       LBV1ClassificationAnswer.from_common(answer)
                        for answer in checklist.answer
                    ],
                    **extra)
@@ -68,25 +65,14 @@ class LBV1Dropdown(LBV1Feature):
     answer: List[LBV1ClassificationAnswer]
 
     def to_common(self) -> Dropdown:
-        return Dropdown(answer=[
-            ClassificationAnswer(feature_schema_id=answer.schema_id,
-                                 name=answer.title,
-                                 extra={
-                                     'feature_id': answer.feature_id,
-                                     'value': answer.value
-                                 }) for answer in self.answer
-        ])
+        return Dropdown(answer=[answer.to_common() for answer in self.answer])
 
     @classmethod
     def from_common(cls, dropdown: Dropdown, feature_schema_id: Cuid,
                     **extra) -> "LBV1Dropdown":
         return cls(schema_id=feature_schema_id,
                    answer=[
-                       LBV1ClassificationAnswer(
-                           schema_id=answer.feature_schema_id,
-                           title=answer.name,
-                           value=answer.extra.get('value'),
-                           feature_id=answer.extra.get('feature_id'))
+                       LBV1ClassificationAnswer.from_common(answer)
                        for answer in dropdown.answer
                    ],
                    **extra)

labelbox/data/serialization/labelbox_v1/label.py

@@ -137,20 +137,17 @@ class LBV1Label(BaseModel):
     label_url: Optional[str] = Extra('View Label')
     has_open_issues: Optional[float] = Extra('Has Open Issues')
     skipped: Optional[bool] = Extra('Skipped')
+    media_type: Optional[str] = Extra('media_type')
 
     def to_common(self) -> Label:
         if isinstance(self.label, list):
             annotations = []
             for lbl in self.label:
                 annotations.extend(lbl.to_common())
-            data = VideoData(url=self.row_data,
-                             external_id=self.external_id,
-                             uid=self.data_row_id)
         else:
             annotations = self.label.to_common()
-            data = self._infer_media_type()
 
-        return Label(data=data,
+        return Label(data=self._data_row_to_common(),
                      uid=self.id,
                      annotations=annotations,
                      extra={
@@ -174,44 +171,49 @@ def from_common(cls, label: Label):
                          external_id=label.data.external_id,
                          **label.extra)
 
-    def _infer_media_type(self):
-        # Video annotations are formatted differently from text and images
-        # So we only need to differentiate those two
+    def _data_row_to_common(self) -> Union[ImageData, TextData, VideoData]:
+        # Use data row information to construct the appropriate annotatin type
         data_row_info = {
+            'url' if self._is_url() else 'text': self.row_data,
             'external_id': self.external_id,
             'uid': self.data_row_id
         }
 
+        self.media_type = self.media_type or self._infer_media_type()
+        media_mapping = {
+            'text': TextData,
+            'image': ImageData,
+            'video': VideoData
+        }
+        if self.media_type not in media_mapping:
+            raise ValueError(
+                f"Annotation types are only supported for {list(media_mapping)} media types."
+                f" Found {self.media_type}.")
+        return media_mapping[self.media_type](**data_row_info)
+
+    def _infer_media_type(self) -> str:
+        # Determines the data row type based on the label content
+        if isinstance(self.label, list):
+            return 'video'
         if self._has_text_annotations():
-            # If it has text annotations then it must be text
-            if self._is_url():
-                return TextData(url=self.row_data, **data_row_info)
-            else:
-                return TextData(text=self.row_data, **data_row_info)
+            return 'text'
         elif self._has_object_annotations():
-            # If it has object annotations and none are text annotations then it must be an image
-            if self._is_url():
-                return ImageData(url=self.row_data, **data_row_info)
-            else:
-                return ImageData(text=self.row_data, **data_row_info)
+            return 'image'
         else:
-            # no annotations to infer data type from.
-            # Use information from the row_data format if possible.
             if self._row_contains((".jpg", ".png", ".jpeg")) and self._is_url():
-                return ImageData(url=self.row_data, **data_row_info)
-            elif self._row_contains(
-                (".txt", ".text", ".html")) and self._is_url():
-                return TextData(url=self.row_data, **data_row_info)
-            elif not self._is_url():
-                return TextData(text=self.row_data, **data_row_info)
+                return 'image'
+            elif (self._row_contains((".txt", ".text", ".html")) and
+                  self._is_url()) or not self._is_url():
+                return 'text'
             else:
-                # This is going to be urls that do not contain any file extensions
-                # This will only occur on skipped images.
-                # To use this converter on data with this url format
-                #   filter out empty examples from the payload before deserializing.
+                #  This condition will occur when a data row url does not contain a file extension
+                #  and the label does not contain object annotations that indicate the media type.
+                #  As a temporary workaround you can explicitly set the media_type
+                #  in each label json payload before converting.
+                #  We will eventually provide the media type in the export.
                 raise TypeError(
-                    "Can't infer data type from row data. Remove empty examples before trying again. "
-                    f"row_data: {self.row_data[:200]}")
+                    f"Can't infer data type from row data. row_data: {self.row_data[:200]}"
+                )
 
     def _has_object_annotations(self):
         return len(self.label.objects) > 0

labelbox/data/serialization/labelbox_v1/objects.py

@@ -27,10 +27,18 @@ def dict(self, *args, **kwargs):
 
     @validator('classifications', pre=True)
     def validate_subclasses(cls, value, field):
-        # Dropdown subclasses create extra unessesary nesting. So we just remove it.
+        # checklist subclasses create extra unessesary nesting. So we just remove it.
         if isinstance(value, list) and len(value):
-            if isinstance(value[0], list):
-                return value[0]
+            subclasses = []
+            for v in value:
+                # this is due to Checklists providing extra brackets []. We grab every item
+                # in the brackets if this is the case
+                if isinstance(v, list):
+                    for inner_v in v:
+                        subclasses.append(inner_v)
+                else:
+                    subclasses.append(v)
+            return subclasses
         return value