kognic-io 1.6.0

Kognic IO Client

Kognic IO Client

Python 3 library providing access to Kognic IO

To install with pip run pip install kognic-io

Documentation & Getting Started Guide

Documentation about how to use the library can be found here

Changelog

All notable changes to this project will be documented in this file.

[1.6.0] - 2023-10-06

Added

• Method to get scenes by uuids, use it with client.scene.get_scenes_by_uuids([...])
• Method to get an annotation by input uuid client.annotation.get_annotation_for_input(input_uuid)
• Method to get all annotations for a scene client.annotation.get_annotations_for_scene(scene_uuid)
• Method to invalidate scenes client.scene.invalidate_scenes([scene_uuid, ...], reason)
• Method to delete an input client.input.delete_input(input_uuid)
• Renamed input_type field on Input to scene_type
• Renamed input_uuid to scene_uuid on the following methods
  • client.annotation.get_annotation
  • client.input.invalidate_inputs
  • client.input.add_annotation_type
  • client.input.remove_annotation_types
  • client.input.get_inputs_by_uuids
• Renamed input_uuid parameter in get_annotation to scene_uuid

Deprecated

• The method invalidate_inputs (use invalidate_scenes instead)
• The method remove_annotation_types (use delete_input instead)
• input_type on the Input model (use scene_type instead)
• input_uuid parameter in favour of scene_uuid on the following methods
  • client.annotation.get_annotation
  • client.input.invalidate_inputs
  • client.input.add_annotation_type
  • client.input.remove_annotation_types
  • client.input.get_inputs_by_uuids

[1.5.0] - 2023-09-29

Added

• Support for asynchronous callbacks in FileData

[1.4.0] - 2023-09-22

Added

• Added missing FileData.Format.WEBP to enable use of WebP for FileData images (which were already supported for InputModel.Image)

Changed

• Renamed lidar_and_cameras attribute to lidars_and_cameras on KognicIOClient
• Renamed uuid field on Input to scene_uuid
• Renamed input_uuid field on Annotation to scene_uuid
• Renamed response from create method from CreateInputResponse to CreateSceneResponse. Also renamed input_uuid field to scene_uuid in the response.
• Renamed InvalidatedReasonInput to InvalidatedReasonScene
• Moved all scene models (for example LidarsAndCameras) from kognic.io.model.input to kognic.io.model.scene

Deprecated (will be removed in a future release)

• lidar_and_cameras attribute on KognicIOClient
• uuid field on Input
• input_uuid field on Annotation
• CreateInputResponse response from create method along with its input_uuid field
• InvalidatedReasonInput
• Scene models in kognic.io.model.input

[1.3.0] - 2023-08-08

Changed

• Validate that all required local image and pointcloud files exist before beginning to create scenes.
  • Validation errors will be raised for missing or empty files.
  • Image and PointCloud filename must refer to a file that exists & contain data unless a callback or raw bytes are provided. (i.e. there is no validation any time FileData are used)
• Support for new camera model named Principal Point Fisheye model.

[1.2.2] - 2023-07-05

Changed

• Pin pydantic version to <2
• Remove dependency on Pillow

[1.2.1] - 2023-05-19

Added

• Add optional attribute max_altitude_angle to fused cylindrical camera model

[1.2.0] - 2023-05-09

Added

• Support for new camera model named Cylindrical model.

Bugfix

• Fixed bug where the calibration types were not used coherently across different calibration models. This resulted in check such as if calibration.calibration_type == CalibrationType.PINHOLE to fail in some cases.

[1.1.9] - 2023-04-12

Added

• Wrapper function create_inputs that allows creation of multiple inputs (with or without a pre-annotation) with a single call. This is useful since it contains wait-logic for the scene to be ready before creating inputs.

[1.1.8] - 2023-03-17

Added

• Support for custom lens projection coefficients for Principal Point Distortion model.

[1.1.6] - 2023-03-14

Added

• Support for new camera model named Fused Cylindrical model.

Changed

• Fixed bug where files intended to be uploaded using data from a callback or in-memory blob would sometimes be incorrectly handled if their callback or blob could be coerced to False.

[1.1.5] - 2023-01-20

Added

• Support for the new input type aggregated_lidars_and_cameras_sequence. This new input type is a sequence of cameras and point clouds, but the point clouds are aggregated over time. More information can be found here.
• New parameter include_content in get_project_annotations that lets you choose whether to fetch the annotation content or not

[1.1.4] - 2022-12-02

Changed

• Bumped dependency on kognic-base-clients to 1.0.3 which introduces retry when uploads fail for TCP or SSL timeout reasons.

[1.1.2] - 2022-11-25

Changed

• Fixed issue with parsing of custom metadata fields

[1.1.1] - 2022-11-11

Added

• The raw bytes for each scene resource (Image, PointCloud) may now be provided in different ways:
  • read from a local file, as before (no changes are needed to continue using this behaviour)
  • passed directly when creating the resource
  • loaded progressively during the upload via a callback

Refer to the updated documentation for details about how to use the new API features.

Changed

• metadata is once again an optional attribute on model classes that use it (CamerasSequence among others)

[1.0.2] - 2022-11-01

Changed

• Bumped dependency on kognic-base-clients to 1.0.1
• Added pagination to the get_inputs method to prevent timeouts

Bugfix

• Metadata should default to an empty dict

[1.0.1] - 2022-10-18

• Updated readme

[1.0.0] - 2022-10-18

This is a major release with breaking changes and refactorings. Please refer to our Annotell-2-Kognic migration guide for details.

Changed

• Annotell becomes Kognic
• Bumped minimum python version from 3.6 to 3.7
• Packaging changes
• Remaining @dataclasses in our model packages are now pydantic models with better validation

Removed

• Deprecated calibration models
• Deprecated get_annotations method

---

Changelog (before rebranding)

[1.3.1] - 2022-09-13

Bugfix

• Changed naming in PrincipalPointDistortionCalibration

[1.3.0] - 2022-09-09

Changed

• Added PrincipalPointDistortionCalibration

[1.2.6] - 2022-08-19

Bugfix

• Fixed error when using deprecated calibration model annotell.input_api.model.calibration.sensors.LidarCalibration
  • We encourage updating to use the newer, typed calibration models in annotell.input_api.model.calibration.{camera, lidar}.*, refer to the calibration documentation for details.

[1.2.5] - 2022-08-18

Changed

• Changed from internalId to sceneUuid on initialize input

Bugfix

• Fixed error when using deprecated calibration model annotell.input_api.model.calibration.sensors.CameraCalibration
  • We encourage updating to use the newer, typed calibration models in annotell.input_api.model.calibration.{camera, lidar}.*, refer to the calibration documentation for details.

[1.2.4] - 2022-07-06

Bugfix

• Fix bug where not specifying feature flags led to a runtime error.

[1.2.3] - 2022-06-23

Added

• Feature flags can be specified when creating inputs. This can be used to disable motion compensation.

[1.2.2] - 2022-06-20

Added

• Ability to specify field of view on LIDAR sensor calibrations.

[1.2.1] - 2022-06-08

Added

• Ability to create inputs from a scene
• Ability to create pre-annotations (limited support)

[1.2.0] - 2022-05-05

Changed

• annotell-base-clients is now a dependency for annotell-input-api, used for http requests and file uploads
• The generator produced by client.annotation.get_project_annotations will now reliably iterate projects with thousands of annotations

Added

• Include annotation_types in Input
• Method to remove annotation types from an input: remove_annotation_types

[1.1.5] - 2022-03-29

Bugfix

• Added 'archived' as ProjectBatchStatus

[1.1.4] - 2022-02-17

Added

• MetaDataContainer added with reserved keyword region.
• Added ImageMetaData for shutter start and end times, used for high accuracy multi-sensor projections

[1.1.3] - 2022-01-05

Bugfix

• Fixed download of annotations

[1.1.2] - 2021-12-03

Changed

• Refactor of file_resource_client.py, split upload and download into separate classes.

Added

• New parameter timeout to InputApiClient, which decides what the timeout in seconds is for calls to Annotell API:s and Google Cloud Storage.
• Retries when a ConnectionError is raised during uploading/downloading of resources to/from Google Cloud Storage.
• New input status Pending. An input will have this status if the input has been validated, but the server is waiting for the associated data to be uploaded. When all data is uploaded the status of the input will change to Processing.

[1.1.1] - 2021-11-11

Changed

• Fixed import statement to work with python < 3.9

[1.1.0] - 2021-11-03

Added

• Two new methods has been added for downloading annotations: client.annotation.get_annotation and client.annotation.get_project_annotations. These two methods will serve annotations in the OpenLABEL format. With this change the previous method for fetching annotations, client.annotation.get_annotations has become deprecated.

• Stricter typing for the calibrations, specifically the camera calibrations. Each of the supported camera calibration models now have their own class in annotell.input_api.model.calibration.camera. Documentation regarding use can be found here: Documentation

• Field of View support for camera calibrations

• New Parameter Xi for Fisheye camera calibration model

Changed

• Two constructor arguments in InputApiClient and FileResourceClient have been renamed from max_upload_retry_attemps, max_upload_retry_wait_time to max_retry_attempts, max_retry_wait_time respectively.

• The old camera calibration class will be deprecated in favour of the new classes

[1.0.8] - 2021-09-07

Added

• A new method has been added, get_inputs_with_uuids, which can fetch inputs using only the input_uuid.
• annoutil has a new flag when fetching inputs, annoutil inputs --uuids <comma_separated_uuids>.
• lidars and lidars_sequence inputs now available through the client.
• A new method has been added, add_annotation_type, which adds additional annotation types to be performed for an input.

Changed

• client.calibration.get_calibration() now properly deserializes calibration into SensorCalibrationEntry instead of keeping it as a dict.

[1.0.7] - 2021-06-11

Added

• created timestamp when querying get_inputs
• Method get_annotation_types

Changed

• input_list_id replaced with annotation_types for all createable resources (Cameras, CamerasSeq, LidarsAndCameras, LidarsAndCamerasSeq).

[1.0.6] - 2021-05-28

Added

• calibration_id now available for created inputs via the client.input.get_inputs method.
• It is now possible to create your project batches on your own using the client.project.create_batch method. Please contact Annotell's Professional Services before using. More information available in the documentation.

[1.0.5] - 2021-05-06

Changed

• Changed the height/width in the unity calibration created in the examples to match the image/videos.
• Added new field in the Input class, view_link. If the Input was successfully created it will contain an URL to view the input in the Annotell app.

Bugfixes

• Fixed issue where invalidate_inputs did not properly discard response content.

[1.0.4] - 2021-04-26

Added

• Added support for providing metadata in the form of a flat KV-pair both on an input-level for all input types, as well as on a frame-level for all sequential input types.

Changed

• Made SensorSpecification Optional for all input types

Removed

• Removed sensor_settings from SensorSpecification. The pixel dimensions are now automatically inferred from videos and images.

[1.0.3] - 2021-04-14

Added

• Added an example for download_annotations
• Added check so that input_list_id and project is not used simultaneously when creating inputs

Changed

• Made client and file_client internal
• Fixed bug where client sometimes didn't raise exception when http calls return error codes
• Bugfix where annoutil didn't work due to missing import
• Clarified examples with different images/videos for different sensors and frames.

Removed

• Removed unnecessary parameters frame_id and relative_timestamp from lidars_and_cameras

[1.0.1] - 2021-04-06

• Use backport of dataclasses to support python 3.6.

[1.0.0] - 2021-03-23

• New major release of client. Reworked to be more internally consistent between input types, and use of project and batch identifiers across methods. See docs for more info.
• client.lidar_and_cameras.create replaces client.create_inputs_point_cloud_with_images
• client.cameras.create replaces client.upload_and_create_images_input_job
• client.annotations.get_annotations replaces client.download_annotations

[0.4.4] - 2021-03-02

• Remove unused dependency on annotell-cloud-storage

[0.4.3] - 2021-02-16

• Fixed import bug in annoutil CLI tool.

[0.4.2] - 2021-02-02

Changed

• Changed url for the get_calibration_data method. Does not affect usage of the method in any way.

[0.4.1] - 2021-01-29

Changed

• Removed unused property deadline from project

[0.4.0] - 2021-01-28

Changed

• Renamed method upload_and_create_images_input_job to create_inputs_images.
• Renamed method list_projects to get_projects.
• Renamed method list_project_batches to get_project_batches.
• Changed behaviour of method download_annotations. The previously optional argumnet request_id has been removed. Additionally, the return signature is changed to return a list of annotations for each input, instead of a dict as before.
• Behaviour of get_inputs has changed. It now receives project (identifier, not numerical id anymore), as well as three optional parameters batch, external_ids and include_invalidated. Returns all inputs belonging to the project, with the option of filtering on batch, external ID and whether or not including invalidated inputs. The returned list of classes had additional fields describing which batch each input belongs to, as well as their status (created, processing, failed, invalidated).
• Changed name of argument input_ids to input_internal_ids for method invalidate_inputs.
• Use backport of dataclasses to support python 3.6.
• Add missing dependency on python-dateutil.

Removed

• Methods count_inputs_for_external_ids, get_internal_ids_for_external_ids, mend_input_data, remove_inputs_from_input_list, list_input_lists, publish_batch, get_requests_for_request_ids, get_requests_for_input_lists, get_input_status, get_input_jobs_status, get_requests_for_project_id, get_datas_for_inputs_by_internal_ids and get_datas_for_inputs_by_external_ids have all been removed.

[0.3.12] - 2021-01-13

Changed

• Removed getting started documentation from README.md and instead link to new docs.

[0.3.11] - 2020-12-14

Changed

• Deserialization bugfix in models for InputBatch and InputBatch.

[0.3.10] - 2020-12-01

Added

• Minor fix in annoutil

[0.3.9] - 2020-11-26

Added

• Bump of required python version to >=3.7
• New explicit models for lidar and camera calibration added.
• publish_batch which accepts project identifier and batch identifier and marks the batch as ready for annotation.

Changed

• Deprecation warning for the old lidar and camera calibration models. No other change in functionality.

[0.3.8] - 2020-11-13

Added

• get_inputs which accepts a project ID or project identifier (external ID) and returns inputs connected to the project. invalidated filter parameter to optionally filter only invalidated inputs. Also exposed in annoutil as annoutil projects 1 --invalidated.

Changed

• invalidate_inputs now accepts annotell internal_ids (UUID) instead of Annotell specific input ids.

[0.3.7] - 2020-11-06

Changed

• bug fix related to oauth session

[0.3.6] - 2020-11-02

Changed

• SLAM - add cuboid timespans, dynamic_objects not includes both cuboids and cuboid_timespans

[0.3.5] - 2020-10-19

Added

• Add support for project and batch identifiers for input request. Specifying project and batch adds input to specified batch. When only sending project, inputs are added to the latest open batch for the project.

Deprecated

• input_list_id will be removed in the 0.4.x version

[0.3.4] - 2020-09-10

Changed

• SLAM - add required sub_sequence_id and optional settings

[0.3.3] - 2020-09-10

Changed

• SLAM - add required sequence_id

[0.3.2] - 2020-09-01

Changed

• SLAM - startTs and endTs not optional in Slam request

[0.3.1] - 2020-07-16

Changed

• If the upload of point clouds or images crashes and returns status code 429, 408 or 5xx the script will retry the upload before crashing. The default settings may be changed when initializing the InputApiClient by specifying values to the max_upload_retry_attempts and max_upload_retry_wait_time parameters.

[0.3.0] - 2020-07-03

Changed

• The method create_inputs_point_cloud_with_images in InputApiClient now takes an extra parameter: dryrun: bool. If set to True all the validation checks will be run but no inputJob will be created, and if it is set to False an inputJob will be created if the validation checks all pass.

Bugfixes

• Fixed bug where the uploading of .csv files to GCS crashed if run on some windows machines.

[0.2.9] - 2020-07-02

Added

• New public method in InputApiClient: count_inputs_for_external_ids.

[0.2.8] - 2020-06-30

Added

• Docstrings for all public methods in the InputApiClient class

[0.2.7] - 2020-06-29

Added

• Require time specification to be send when posting slam requests

[0.2.6] - 2020-06-26

Changed

• Removed CalibrationSpec from CalibratedSceneMetaData and SlamMetaData. Updated so that create_calibration_data in InputApiClient only takes a CalibrationSpec as parameter.

[0.2.5] - 2020-06-22

Bugfixes

• Fixed issue where a path including a "~" would not expand correctly.

[0.2.4] - 2020-06-17

Changed

• Changed pointcloud_with_images model. Images and point clouds are now represented as Image and PointCloud containing filename and source. Consequently, images_to_source is removed from SourceSpecification.

Added

• create Image inputs via create_images_input_job
• It's now possible to invalidate erroneous inputs via invalidate_inputs
• Support for removing specific inputs via remove_inputs_from_input_list
• SLAM support (not generally available)

Bugfixes

• Fixed issue where annoutils would not deserialize datas correctly when querying datas by internalId

[0.2.3] - 2020-04-21

Changed

• Changed how timestamps are represented when receiving responses.

[0.2.2] - 2020-04-17

Added

• Methods get_datas_for_inputs_by_internal_ids and get_datas_for_inputs_by_external_ids can be used to get which Data are part of an Input, useful in order to check which images, lidar-files have been uploaded. Both are also available in the CLI via :

$ annoutil inputs --get-datas <internal_ids>
$ annoutil inputs-externalid --get-datas <external_ids>

• Support has been added for Kannala camera types. Whenever adding calibration for Kannala undistortion coefficients must also be added.
• Calibration is now represented as a class and is no longer just a dictionary, making it easier to understand how the Annotell format is structured and used.

[0.2.0] - 2020-04-16

Changed

• Change constructor to disable legacy api token support and only accept an auth parameter

[0.1.5] - 2020-04-07

Added

• Method get_input_jobs_status now accepts lists of internal_ids and external_ids as arguments.