Annotation Types
Annotation Types are something that can be configured for a project either during or after the Guideline Agreement Process (GAP). During the GAP a set of taxonomies are defined which will be used in order to annotate all the inputs that are uploaded to the platform.
Annotation Types are identifiers that each map to one of these taxonomies and allows you to programmatically define, for every input that you create, what taxonomies that you would like that input to be annotated with. You do not need to use annotation types in a project but by skipping annotation types your inputs would always be created for all taxonomies available in the specified batch.
It is possible to add additional Annotation Types after a project is created, and make them available to future batches that you create. If you would like Annotation Types to be added to your project please get in touch with Kognic's Professional Services team.
Examples
For the following examples we will be creating cameras_sequence inputs, however the
procedure would be identical for any other input type. We will also assume that the
project example_project_id is configured with the Annotation Types:
static_objects, and dynamic_objects, and that they are also available in the batch
example_batch_id.
Get Annotation Types for Project
from kognic.io.client import KognicIOClient
client = KognicIOClient()
project_annotation_types = client.project.get_annotation_types(project="example_project_id")
This will return a list of all Annotation Types available in the project.
Get Annotation Types for a specified Project Batch
from kognic.io.client import KognicIOClient
client = KognicIOClient()
batch_annotation_types = client.project.get_annotation_types(project="example_project_id",
batch="example_batch_id")
This will return a list of all Annotation Types available in the specified batch. Note that this list does not need to contain all Annotation Types in the project.
Create inputs for specific Annotation Types
from kognic.io.client import KognicIOClient
from kognic.io.model.scene.cameras_sequence import CamerasSequence
client = KognicIOClient()
camera_input = CamerasSequence(external_id=..., frames=...)
client.cameras_sequence.create(camera_input,
project="example_project_id",
batch="example_batch_id",
annotation_types=["static_objects", "dynamic_objects"])
The above example will create a new input which will be annotated for the annotation types specified. If one or more of the specified annotation types would not be available in the specified batch the validation in the API would fail.
In these examples we have specified which batch the inputs should be created for, but this is optional. If
no batch is specified the inputs will be created in the latest batch with status open.
Create inputs for all Annotation Types in batch
from kognic.io.client import KognicIOClient
from kognic.io.model.scene.cameras_sequence import CamerasSequence
client = KognicIOClient()
camera_input = CamerasSequence(external_id=..., frames=...)
client.cameras_sequence.create(camera_input,
project="example_project_id",
batch="example_batch_id")
The above example will create a new input which will be annotated for all Annotation
Types available in the batch example_batch_id. However, this way it is not explicit
what Annotation Types that the inputs will be annotated with, and you would not get an
error if e.g. static_objects was missing from the specified batch.
In order to get the best possible validation it is recommended that you always specify annotation types when you create inputs.
Add/remove annotation types for an input (deprecated)
Removing annotation types has been deprecated in favor of Deleting Inputs and will be removed in the near future.
Adding annotation types has been deprecated in favor of Creating Inputs from Scene and will be removed in the near future.
Adding an annotation type to an input means that an annotation will be produced for that input with the specified annotation type. In the same way, removing annotation types from an input means that annotations will not be produced for that input with the specified annotation types.
In the case when multiple annotation types are annotated in the same task, it is enough to specify one annotation type when adding but all annotation types must be specified when removing.
Note that it is currently not possible to add an annotation type that has already been removed from an input.
from kognic.io.client import KognicIOClient
client = KognicIOClient()
input_uuid = 'cca60a67-cb68-4645-8bae-00c6e6415555'
# Add an annotation type to an input
client.input.add_annotation_type(input_uuid=input_uuid, annotation_type="annotation-type")
# Remove annotation types from an input
annotation_types = ["annotation-type-1", "annotation-type-2", ...]
client.input.remove_annotation_types(input_uuid=input_uuid, annotation_type=annotation_types)