SceneEngineClient

class scenebox.clients.SceneEngineClient(auth_token=None, scene_engine_url=None, asset_manager_url=None, environment=None, artifacts_path='.', kv_store=None, num_threads=20)

Acts as the root client for all user-tasks on SceneBox.

Create a SceneEngineClient object. Used to interact with SceneBox’s high-level REST APIs.

Parameters:
  • auth_token – Personal token associated with a valid SceneBox account. Find this token under user profile on the SceneBox web app.

  • scene_engine_url – URL for the SceneEngine REST server.

  • asset_manager_url – URL for the AssetManager associated with the SceneEngine server.

  • environment – The environment that the client is going to use (dev or prod). It will override

  • artifacts_path – SceneBox’s output directory.

  • kv_store – key-value store to use in asset-manager-client to cache bytes. max_threads:

    maximum number of threads to run internal processes in. Default specified in constants file.

Methods

add_annotation

Add a single annotation.

add_annotations

Add several annotations at once.

add_assets_to_set

Add assets to a set.

add_auxiliary_session_data

Add auxiliary session data e.g.

add_coco_annotations_from_folder

Add several images and/or COCO annotations.

add_comments

Add a comment to a time segment of a session.

add_df

Add a measurement DataFrame to a session as scalar entities.

add_ego_pose

Add an ego pose file to a session. An ego pose file has a series of frames in which a list of timestamped objects with their ego position in “lat” and “lon” is given. Example: [ { “frame”:0, “timestamp”:”2021-08-24T21:15:15.574000+00:00”, “objects”:{ “0”:{ “lat”:33.6, “lon”:-0.15 }, “1”:{ “lat”:5.6, “lon”:-3.45 }, “2”:{ “lat”:23.25, “lon”:-3.85 }, “3”:{ “lat”:83.0, “lon”:-5.9 }, “4”:{ “lat”:198.55, “lon”:12.15 } } }, { “frame”:1, “timestamp”:”2021-08-24T21:15:15.599000+00:00”, “objects”:{ “5”:{ “lat”:33.55, “lon”:-0.15 }, “6”:{ “lat”:5.7, “lon”:-3.45 }, “7”:{ “lat”:23.25, “lon”:-3.85 }, “8”:{ “lat”:82.45, “lon”:-5.9 }, “9”:{ “lat”:197.8, “lon”:12.15 } } } ].

add_embeddings

Add existing embeddings to an asset.

add_enrichments_configs

Add enrichment configurations to session metadata.

add_entities

Add a list of entities to the specified sessions.

add_entity

Add an entity to the specified session.

add_event_interval

Add an event interval to a session.

add_event_intervals

Add a list of several event intervals to a session.

add_geolocation_entities

Add geolocation entities to a session.

add_geolocation_entity

Add a geolocation entity to a session.

add_image

Upload an image onto SceneBox.

add_images

Upload multiple images onto SceneBox.

add_images_from_folder

Upload images from a single local folder path onto SceneBox.

add_lidar

Upload a LIDAR onto SceneBox.

add_lidars

Upload multiple videos onto SceneBox.

add_organization

Add a new SceneBox organization.

add_point_events

Add a list of several point events to a session.

add_rosbag2

Upload Rosbag2 onto SceneBox.

add_scalar_intervals

Add a timeseries with interval values to a session.

add_scalar_measurements

Add a timeseries with point values (“scalars”) to a session.

add_session

Add a session.

add_sets_to_project

Add sets to a project.

add_source_data

Add source data to session metadata.

add_task_to_campaign

Add new tasks to a campaign.

add_timeseries_csv

Add a timeseries from a CSV file.

add_user

Add a new SceneBox user account to your organization.

add_video

Upload a video onto SceneBox.

add_videos

Upload multiple videos onto SceneBox.

annotations_to_objects

Extracts objects from annotations.

assets_in_set

List assets within a set.

cleanup_assets_annotation_meta

Cleanup annotation lists by removing repeated and non-existent entries in an asset’s metadata. asset_ids: List of asset ids to clean-up. asset_type: The type of asset. To ensure using a valid asset_type, you can import AssetsConstants from scenebox.constants.

clear_cache

Clears scene engine’s Redis cache by organization Args: all_organizations: if True, Redis cache for all organization is cleared. Requires Admin privilege partitions: list of partitions (like images, sets, annotations) to clear their cache, if not set, all is cleared.

compare_sets

Compare sets based on a given metric_type in an embedding space with given index_name.

compress_images

Compress a list of images.

compress_lidars

Compress a list of videos.

compress_videos

Compress a list of videos.

concatenate_videos

Concatenate a list of videos into a single video.

count

Retrieve summary of asset metadata with a search query.

create_annotation_instruction

Create an annotation instruction.

create_campaign

Create a Campaign A Campaign is a collection of tasks to run on the raw data or metadata on SceneBox.

create_config

Create a config.

create_operation

Create an operation inside a project.

create_project

Create a project.

create_set

Create a named dataset.

create_umap

Applies UMAP onto existing asset indices.

delete_annotation_instruction

Delete an existing annotation instruction.

delete_annotations

Delete annotations using a list of ids or a search query, and updates metadata of corresponding assets.

delete_campaign

Delete a campaign Deleting a campaign will delete all tasks and by extension all operation runs contained in it.

delete_campaign_operation_config

Delete a config.

delete_config

Delete a config.

delete_operation

Delete an operation.

delete_project

Delete a project.

delete_session

Delete an existing session.

delete_set

Delete a set.

delete_task

Delete a task Deleting a task will delete operation instances associated with it.

disable_webapp_features

Disable a list of features on SceneBox WebApp.

display_dashboard

Display a given dashboard on the SceneBox web app.

display_image

Display a given image on the SceneBox web app.

display_lidar

Display a given Lidar on the SceneBox web app.

display_object

Display a given object on the SceneBox web app.

display_projects

Display a given project on the SceneBox web app.

display_session

Display a given session on the SceneBox web app.

display_set

Display a given set on the SceneBox web app.

display_video

Display a given image on the SceneBox web app.

download_annotations

Download annotations to a destination folder including the mask files

download_timeseries

Download the timeseries using a search query.

download_zipfile

Download a zipfile into either a bytesIO object or a file.

enable_webapp_features

Enable a list of features on SceneBox WebApp.

entity_summary

Get the entity summary for a session.

extend_time_span

Extend the start time and end time of a session based on min/max of a list of timestamps.

extract_frame_thumbnails

Extract thumbnails for a list of frames in a video.

extract_images_from_videos

Extract images from video

extract_subclip_from_videos

Create a subclip of a video.

flush_entities_buffer

write all of the entities in the buffer.

get_annotation_labels

Returns all labels for all images or images that satisfy a search query

get_annotation_sources

Returns all sources of annotation for all images or images that satisfy a search query

get_asset_manager

Set asset_manager state.

get_campaign_operation_config

Get config metadata.

get_config

Get config metadata.

get_job

Get the job_manager_client.Job object associated with the provided job ID.

get_job_manager

Set asset_manager state.

get_metadata

Get the metadata of an asset.

get_operation

Get specifications of an operation

get_operation_status

Get the status of an operation.

get_performance_metrics

Get performance metrics.

get_performance_metrics_per_dimension

Get performance metrics per dimension.

get_platform_version

Get version of the SceneBox Client.

get_searchable_field_statistics

Get the metadata searchable field statistics for a certain asset type.

get_session_resolution_status

Get the status of the session resolution task (if it exists).

get_set_status

Get the status of a set.

get_sets_comparison_matrix

Get the comparison matrix of an index.

get_streamable_set

A StreamableSet is a SceneBox object that can be used to build dataloaders for a set.

get_video_frame_thumbnail

Get thumbnail png for requested frame fram a video.

image_properties_enrichment

Enrich images or objects with classical image processing features.

index_recording

Index Recording.

index_rosbag

Index Ros1 files.

index_rosbag2

Index Ros2 files.

index_s3_images_batch

Index s3 images.

is_coco_annotation_file_valid

Tests if a COCO annotation file is valid.

jira_create_issue

create a Jira issue

jira_is_connected

Check Jira connection :returns: is jira connected :rtype: bool

jira_link_new_issue_to_existing

create a new Jira issue and link to an existing one

list_operations

Get specifications of an operation

list_organizations

Returns a list of all organizations.

list_ownerships

Returns a list of all ownerships.

list_users

Returns a list of all users.

make_video_from_image_uris

Make video from frames in a given cloud_storage, bucket, and folder.

make_video_with_annotations_from_uris

Make video with overlayed annotations from provided image uris and annotaion uris per provider

model_inference

Perform inference from a list of supported models.

modify_organization

Modify an existing organization. :Parameters: * organization_id – Organization id * organization_name – The new organization name.

modify_user

Modify an existing user account.

register_rosbag_session

Register a rosbag session with the SceneEngine.

remove_annotation_sources

Delete annotations using a list of ids or a search query, and updates metadata of corresponding assets.

remove_assets_from_set

Remove assets from a set either by query or by IDs.

remove_sets_from_project

Remove sets from a project.

resolve_session

Resolve a session.

run_operation_instance

Run an individual instance of an operation contained in a task.

save_campaign_operation_config

Create a config for a campaign operation.

search_assets

Retrieve asset IDs with a search query.

search_campaigns

Search for campaigns that satisfy the filtering criteria

search_meta

Retrieve asset metadata with a search query.

search_similar_assets

Find top k similar assets to an existing asset with embeddings.

search_tasks

Search for tasks that satify the filtering criteria

search_within_set

Search for assets within a set.

set_is_locked

Check the lock status of a set.

set_lock

Lock a set.

set_unlock

Unlocks a set.

similarity_search_bulk_index

Bulk indexes embeddings to allow for similarity search.

slice_session

slice an existing session with start and end time.

split_set

Split a set into subsets by randomly sampling and distributing constituent assets.

start_operation

Start an operation.

summary_meta

Retrieve summary of asset metadata with a search query.

update_artifact_path

Updates the SceneEngineClient artifact path.

update_time_span

Update the start/end times of a session.

whoami

Get logged-in user authentication details

with_auth

Save auth_token into SceneEngineClient as an instance attribute.

zip_set

Create a zipfile of a set.

zip_set_and_download

Zip and download an existing set.

whoami()

Get logged-in user authentication details

Returns:

A dictionary with the following keys: “version”,”build_host””token”,”username”,”role”,”organization”

Return type:

dict

with_auth(auth_token=None)

Save auth_token into SceneEngineClient as an instance attribute. Authorizes the current SceneEngineClient object with the auth_token provided in this call or uses the auth_token the client was created with.

Parameters:

auth_token – Personal token associated with a valid SceneBox account. Find this token under user profile on the SceneBox web app.

get_platform_version()

Get version of the SceneBox Client.

Returns:

SceneBox version number

Return type:

str

Raises:

ValueError – If a valid version cannot be obtained from SceneBox.

update_artifact_path(artifacts_path)

Updates the SceneEngineClient artifact path.

Parameters:

artifacts_path – Path that specifies the location to save output files.

get_job(job_id)

Get the job_manager_client.Job object associated with the provided job ID.

Can be a currently executing or finished Job.

Parameters:

job_id – The job ID associated with the Job of interest.

Returns:

Job object with the job id job_id.

Return type:

Job

create_project(name, tags=None)

Create a project.

Projects are used to organize and create data workflows. Once a project is made, users can add sets, and apply operations associated with object/embeddings/annotation extraction, and more.

Parameters:
  • name – Name for the created project.

  • tags – A list of tags to associate this project with. Helpful for filtering them later.

Returns:

The ID of the created project.

Return type:

str

Raises:

ResponseError – If the project could not be created.

delete_project(project_id)

Delete a project.

Delete a project. Deleting a project will not delete any related sets (primary or curated) or annotation recipes.

Parameters:

project_id – The name of the project to delete.

Raises:

ResponseError – If the project could not be deleted.

create_operation(name, operation_type, project_id, config, stage=0, description=None)

Create an operation inside a project.

Operations are the custom workflows that can be created inside projects. Several operations can be added to a single project. Choose from annotation, dataset similarity search, consensus, object extraction, and model inference operation types.

Parameters:
  • name – The name of the operation to be created.

  • operation_type – The type of the operation to execute. Must match a constant in scenebox.constants.OperationTypes.VALID_TYPES.

  • project_id – The ID of the project to add the operation to.

  • config – Configurations for the operation to be added. If operation_type is "annotation", then config must contain a dict entry with a key named “annotation_instruction_id” with the value of the relevant annotation instruction details.

  • stage – The stage of the operation to be added. Represents the order of the operations to be executed inside a project.

  • description – A brief description outlining the purpose of the operation, what the operation will accomplish, etc.

Returns:

The ID of the created operation.

Return type:

str

Raises:
  • ResponseError: – If the project could not be created.

  • OperationError: – If the provided operation type is not a valid operation type.

delete_operation(operation_id)

Delete an operation.

Delete an operation from a project. Deleting an operation will not delete any related sets (primary or curated) or annotation recipes used by or generated by the operation.

Parameters:

operation_id – The ID of the operation to delete.

Raises:

ResponseError – If the operation could not be deleted.

start_operation(operation_id, operation_step=None, wait_for_completion=False)

Start an operation.

Parameters:
  • operation_id – The ID of the operation to execute.

  • operation_step – If the operation is a CVAT or SuperAnnotate operation, chooses whether to send data to the given annotator, or receive annotations. To send data to the annotator, pass send_data. To receive, input receive_annotations. Otherwise, has no effect.

  • wait_for_completion – If True, polls until job is complete. Otherwise, continues execution and does not raise an error if the job does not complete.

Returns:

The job ID of the Job that starts the operation.

Return type:

str

get_operation_status(operation_id)

Get the status of an operation.

Parameters:

operation_id – The ID of the operation to execute.

Returns:

The full status of the operation.

Return type:

dict

add_sets_to_project(project_id, sets, stage=0)

Add sets to a project.

Specify sets to add to a project. To run a given operation (associated with some project) on a set, the set must be added to the same project as the desired operation.

Parameters:
  • project_id – The ID of the project to add the given sets to.

  • sets – List of IDs of sets to add to the project.

  • stage – The stage associated with the set. Match this with the stage of the desired operation to run.

remove_sets_from_project(project_id, sets)

Remove sets from a project.

Parameters:
  • project_id – ID of the project to remove sets from.

  • sets – A list of dictionaries listing sets to be removed. Dict format:

    [{“set_id”: id_of_the_set_tobe_removed, “stage”: “project_stage_for_the_set”}]

create_set(name, asset_type, origin_set_id=None, expected_count=None, tags=None, id=None, aux_metadata=None, is_primary=False, raise_if_set_exists=False)

Create a named dataset.

Set creation is helpful for performing data operations on several assets at once (e.g. ingesting an entire set of images/videos/LIDARs/etc. into SceneBox). Each set can only contain one asset type. Can optionally raise a SetsError if a set by the given name already exists.

Parameters:
  • name – The name to give to the set.

  • asset_type – The type of assets for the set to contain. e.g. images, videos, LIDARs, etc. To ensure using a valid asset_type, you can import AssetsConstants from scenebox.constants

  • origin_set_id – Id of the set, the new set is created from. Useful when creating a set by deriving its assets from other sets.

  • expected_count – Estimated count of the number of assets this set is expected to contain.

  • tags – Labels associated with the set. Allows for easy set search.

  • id – Optional unique id for set (needs to be a string)

  • aux_metadata – Optional dictionary of key and values to be added to sets metadata

  • is_primary – If true, Gives the set an additional “Primary” tag for easy set sorting. Otherwise, does nothing.

  • raise_if_set_exists – Raises a SetsError if the name parameter matches an existing set.

Returns:

The name of the successfully created set.

Return type:

str

Raises:

SetsError – If raise_if_set_exists is True, then will raise an error if the given set name matches an existing set.

create_annotation_instruction(name, key, annotator, annotation_type, payload, media_type='images', description=None)

Create an annotation instruction.

Annotation instructions hold the configurations for annotating assets with external annotators. When executing an operation of type “annotation”, the relevant annotation instruction card is parsed.

Parameters:
  • name – Name of the annotation instruction to create.

  • key – Authentication key to send to Scale or SuperAnnotate.

  • annotator – Name of the annotator to use. Choose an option from scenebox.constants.AnnotationProviders.VALID_PROVIDERS.

  • annotation_type – Name of the annotation type to use. Choose an option from scenebox.constants.AnnotationTypes.VALID_TYPES.

  • payload – Configuration for the given annotator.

  • media_type – Media type to send to the annotator. Choose an option from scenebox.constants.AssetConstants.VALID_TYPES.

  • description – A brief description of the annotation instruction’s purpose, what it aims to accomplish, etc.

Returns:

The ID of the created annotation instruction.

Return type:

str

Raises:

ResponseError: – If the annotation instruction could not be created.

delete_annotation_instruction(annotation_instruction_id)

Delete an existing annotation instruction.

Parameters:

annotation_instruction_id – The ID of the annotation instruction to delete.

Raises:

ResponseError: – If the annotation instruction could not be deleted.

create_config(name, type, config, description=None, id=None, **kwarg)

Create a config.

Parameters:
  • name – Name of the configuration.

  • type – Type of the config to create. Choose an option from scenebox.constants.ConfigTypes.VALID_TYPES.

  • config – Body of the config to create. Form dependent on the configuration needs.

  • description – A brief description of the config’s purpose, settings, etc. Same as the payload argument for self.create_annotation_instruction.

  • id – optional unique identifier

  • **kwarg – Any other misc. kwargs to save into the config.

Returns:

The ID of the created config.

Return type:

str

delete_config(config_id)

Delete a config.

Parameters:

config_id – The ID of the config to delete.

Raises:

ResponseError: – If the config could not be deleted.

get_config(config_id, complete_info=False)

Get config metadata.

Parameters:
  • config_id – The ID of the config to receive the metadata of.

  • complete_info – If True, returns the the entire metadata of the config (with the config parameters contained inside). Otherwise, only returns the config parameters.

Returns:

The metadata and/or parameters of the desired config.

Return type:

dict

Raises:

ResponseError: – If the config metadata could not be obtained.

set_lock(set_id)

Lock a set.

Locks a set from changes. Useful for protecting a set (i.e. marks a set as “in use”) when performing a data operation on a set’s assets.

Parameters:

set_id – The ID of the set to lock.

Returns:

Always returns True.

Return type:

bool

Raises:

HTTPError: – If a server or client error occurs.

set_unlock(set_id)

Unlocks a set.

Releases a set from protection/external changes.

Parameters:

set_id – The ID of the set to unlock.

Returns:

Always returns True.

Return type:

bool

Raises:

HTTPError: – If a server or client error occurs.

set_is_locked(set_id)

Check the lock status of a set.

Parameters:

set_id – The ID of the set to check the lock status of.

Returns:

Returns True if the set is locked. Otherwise, returns False if the set is not locked.

Return type:

bool

Raises:

HTTPError: – If a server or client error occurs.

delete_set(set_id, delete_content=False, wait_for_completion=False)

Delete a set.

Parameters:
  • set_id – The ID of the set to delete.

  • delete_content – Should the content of the set be deleted or just the set itself. Set to False by default meaning the content wont be deleted.

  • wait_for_completion – If True, polls until job is complete. Otherwise, continues execution and does not raise an error if the job does not complete.

Returns:

The ID of the Job that attempts to delete the set.

Return type:

str

split_set(set_id, percentages_or_sections, subset_names=None, random_seed=42, wait_for_completion=False)

Split a set into subsets by randomly sampling and distributing constituent assets.

Parameters:
  • set_id – The ID of the parent set to split.

  • percentages_or_sections – percentages_or_sections is an integer n or a list of floating point values. If percentages_or_sections is an integer n, the input set is split into n sections. If input set size is divisible by n, each section will be of equal size, input_set_size / n. If input set size is not divisible by n, the sizes of the first int(input_set_size % n) sections will have size int(input_set_size / n) + 1, and the rest will have size int(input_set_size / n).

    If percentages_or_sections is a list of floating point values, the input set is split into subsets with the amount of data in each split i as floor(percentage_or_sections[i] * input set size). Any leftover data due to the floor operation is equally redistributed into len(percentage_or_sections) subsets according to the same logic as with percentages_or_sections being an integer value.

  • subset_names – Names for the subsets to be created. Subsets corresponding to the provided subset names should not previously exist.

  • random_seed – Optional. Set a different seed to get a different random subsample of the data in each split.

  • wait_for_completion – If True, polls until job is complete. Otherwise, continues execution and does not raise an error if the job does not complete. If numof splits is greater than 5, wait_for_completion cannot be set to False.

Returns:

The IDs of subsets created.

Return type:

List[str]

get_set_status(set_id)

Get the status of a set.

Parameters:

set_id – The ID of the set to get the status of.

Returns:

The full status of the desired set.

Return type:

dict

Raises:

HTTPError: – If a server or client error occurs.

add_assets_to_set(set_id, search=None, ids=None, limit=None, wait_for_availability=True, timeout=30, wait_for_completion=False)

Add assets to a set.

Add assets to a set either with a search query, or by IDs. Utilize only one of

Parameters:
  • search – Query to locate the data subset of interest. Filters through existing assets according to the dictionary passed.

  • ids – IDs of the assets to add to the given set.

  • set_id – The ID of the set to add assets to.

  • limit – The limit of the number of additions. If the limit is reached, randomly down-sample the list of assets to add.

  • wait_for_availability – If True and the specified set is locked, waits until timeout for the set to unlock. Otherwise, if False, raises a SetsError if the set is locked.

  • timeout – If wait_for_availability is True, represents the maximum amount of time for a set to unlock. Otherwise, has no effect.

  • wait_for_completion – If True, polls until job is complete. Otherwise, continues execution and does not raise an error if the job does not complete.

Returns:

ID of the Job that runs/ran the add assets job.

Return type:

str

Raises:
  • SetsError – If wait_for_availability is False, and the specified set is locked. If neither ids or search are passed.

  • TimeoutError – If wait_for_availability is True, and timeout is reached before the desired set is unlocked.

  • ResponseError – If assets could not be added to the desired set.

remove_assets_from_set(search=None, ids=None, set_id=None, wait_for_availability=True, timeout=30, wait_for_completion=False)

Remove assets from a set either by query or by IDs.

Parameters:
  • search – Query to locate the data subset of interest. Filters through existing assets according to the dictionary passed.

  • ids – IDs of the assets to add to the given set.

  • set_id – The ID of the set to add assets to.

  • wait_for_availability – If True and the specified set is locked, waits until timeout for the set to unlock. Otherwise, if False, raises a SetsError if the set is locked.

  • timeout – If wait_for_availability is True, represents the maximum amount of time for a set to unlock. Otherwise, has no effect.

  • wait_for_completion – If True, polls until job is complete. Otherwise, continues execution and does not raise an error if the job does not complete.

Returns:

The ID of the job that runs/ran the Job to remove assets from a set.

Return type:

str

assets_in_set(set_id=None)

List assets within a set.

Parameters:

set_id – The ID of the set to list assets from.

Returns:

A list of the IDs of the assets contained in the specified set.

Return type:

List[str]

search_within_set(set_id, search=None, show_meta=True)

Search for assets within a set.

Parameters:
  • search – Search query to send to the endpoint. Filters through existing sets according to the dictionary passed.

  • set_id – The ID of the set to search.

  • show_meta – If True, returns the metadata of each asset in the set. Otherwise, simply lists the IDs of each asset in the set.

Returns:

Contains asset IDs and/or metadata for the assets in the given set.

Return type:

List

Raises:

ResponseError – If the search could not be performed.

zip_set(set_id)

Create a zipfile of a set.

Parameters:

set_id – The ID of the set to zip.

Returns:

The ID of the job that zipped the set, The zipfile ID of the zipped set).

Return type:

(str, str)

Raises:

ValueError – If no set ID is provided.

compare_sets(sets, index_name, metric_type=None, metric_params=None, wait_for_completion=True)

Compare sets based on a given metric_type in an embedding space with given index_name.

Parameters:
  • sets – The IDs of the sets to compare with each other.

  • index_name – The name of the embedding space.

  • metric_type – The metric type that the comparison is based on; if not given, default value is wasserstein.

  • metric_params – key-values of any parameters that are needed to calculate the metric; if not given, default values for num_projections and num_seeds are 25 and 50, respectively, for wasserstein distance calculation.

  • wait_for_completion – If True, polls until job is complete. Otherwise, continues execution and does not raise an error if the job does not complete.

Returns:

The ID of the job that compared sets.

Return type:

str

Raises:

ValueError – If sets is empty or contains only one set id. If index_name is not provided.

get_sets_comparison_matrix(search=None, index_name=None)

Get the comparison matrix of an index.

Parameters:
  • search – Query to locate the data subset of interest. Filters through existing assets according to the dictionary passed.

  • index_name – index_name of the embeddings of interest.

Returns:

The requested comparison matrix.

Return type:

List[dict]

get_performance_metrics(comparison_task_id, search=None, iou_threshold=0.0)

Get performance metrics.

Get performance metrics including precision, recall, and f1 score for a given search on annotations comparisons asset space and iou threshold.

Parameters:
  • comparison_task_id – id of the comparison task to calculate metrics for

  • search – Query to locate the data subset of interest. Filters through existing annotations comparisons according to the dictionary passed.

    # Search example
    search = {
                "filters": [{"field": "sets", "values": ["some_set_id"]}]
             }
    
  • iou_threshold – iou threshold to use in confusion matrix calculation

Returns:

The requested performance metrics.

# Example return
{
    "cat": {'precision': 0.75, 'recall': 0.47368421052631576, 'f1': 0.5806451612903226},
    "dog": {'precision': 0, 'recall': 0, 'f1': N/A}
}

Return type:

Dict[str, Any]

get_performance_metrics_per_dimension(comparison_task_id, dimension, iou_threshold=0.0, search=None, force_recalculation=False)

Get performance metrics per dimension.

Get performance metrics including precision, recall, and f1 score per existing buckets in a given dimension e.g. image ids.

Parameters:
  • comparison_task_id – id of the comparison task to calculate metrics for

  • dimension – the dimension for which the performances should be calculated e.g. “id”

  • iou_threshold – iou threshold to use in confusion matrix calculation

  • search – Query images to locate the data subset of interest.

    # Search example:
    search = {
                "filters": [{"field": "sensor", "values": ["camera_1"]}]
             }
    
  • force_recalculation – If True, cached results are ignored if available.

Returns:

A dictionary that has 2 keys; job_id, and performances.

The job_id key is mapped to the ID of the job that calculated the performances, and the performances key is mapped to the resultant performance.

# Example return
{
    "job_id": "some_job_id",
        "performances": {
            "image_id_1": {
                "cat": {'precision': 0.75, 'recall': 0.47368421052631576, 'f1': 0.5806451612903226},
                "dog": {'precision': 0, 'recall': 0, 'f1': N/A}
            },
            "image_id_2": {
                "cat": {'precision': 0.25, 'recall': 0.5, 'f1': 0.33333333333},
                "dog": {'precision': 0.5, 'recall': 1.0, 'f1': 0.66666666666}
            }
        }
    }

Return type:

Dict[str, Dict[str, Any]]

download_zipfile(zip_id, output_filename=None)

Download a zipfile into either a bytesIO object or a file.

Parameters:
  • zip_id – The ID of the zipfile to download.

  • output_filename – The filename to give the downloaded zipfile.

Returns:

If output_filename is provided, returns the path where the zipfile was locally saved as str. Otherwise, the zipfile is returned directly as a io.BytesIO object.

Return type:

Union[str, io.BytesIO]

download_timeseries(search_dic=None, fields=None)

Download the timeseries using a search query.

Parameters:
  • search_dic – Query to locate the data subset of interest. Filters through existing assets according to the dictionary passed.

  • fields – Field names of fields to return from each session.

Returns:

Mapping between session uids, and the returned timeseries data.

Return type:

dict

zip_set_and_download(set_id=None, filename=None)

Zip and download an existing set.

The downloaded zip file contains several subfolders/files encapsulating the logged data on SceneBox. More specifically, this includes the following: - Asset files with their original extension (images/videos/LIDARs, etc.) - Asset metadata in .json format - Annotation data in .json format (if available)

Parameters:
  • set_id – Set name of the set to zip and download

  • filename – Name of the downloaded zip folder

Returns:

If filename is provided, returns the path where the zipfile was locally saved as str. Otherwise, the zipfile is returned directly as a io.BytesIO object.

Return type:

Union[str, io.BytesIO]

Raises:

ValueError – If no set_id is provided.

get_asset_manager(asset_type)

Set asset_manager state.

Set the asset_manager state to access particular assets. This is often used in chaining: Eg. client.with_asset_state(“images”, True).search_files({})

Parameters:

asset_type – Desired AssetManagerClient asset type. To ensure using a valid asset_type, you can import AssetsConstants from scenebox.constants

Returns:

The AssetManagerClient associated with the specified asset.

Return type:

AssetManagerClient

get_job_manager(job_id=None)

Set asset_manager state.

Set the asset_manager state to access particular assets. This is often used in chaining: Eg. client.with_asset_state(“images”, True).search_files({})

Parameters:

job_id – Optional. A job id for the job to manage. If not provided, a job manager is created with a guid for job id.

Returns:

The AssetManagerClient associated with the specified asset.

Return type:

AssetManagerClient

get_streamable_set(set_id, force_recreate=False, shard_size=50000)

A StreamableSet is a SceneBox object that can be used to build dataloaders for a set. A dataloader enables fetching batches of tuples of data such as (images, annotations) from scenebox that can be used to in a machine learning models’s training or testing loop.

Parameters:
  • set_id – String. ID of the set that you want to stream.

  • force_recreate – Boolean. Default False. Flag for mandatory update of the StreamableSet object’s data repository.

  • shard_size – Integer. Maximum number of assets in a shard. Data will be split into multiple shards of shard_size.

Returns:

A StreamableSet object that supports pytorch dataloader methods.

Return type:

StreamableSet

register_rosbag_session(session_name, session_directory_path=None, rosbag_ids=None, config_name=None, wait_for_completion=True)

Register a rosbag session with the SceneEngine.

Parameters:
  • session_name – Name of the session.

  • session_directory_path – Local path (on ros-extractor) to a session directory.

  • rosbag_ids – List of rosbag IDs belonging to the session.

  • config_name – Name of the session configuration file.

  • wait_for_completion – If True, polls until job is complete. Otherwise, continues execution and does not raise an error if the job does not complete.

Returns:

the Job ID that carried out the indexing job, and the session UID of the created session.

Return type:

Tuple[str, str]

get_session_resolution_status(session_uid)

Get the status of the session resolution task (if it exists).

Parameters:

session_uid – The session uid to get the session resolution status of.

Returns:

The session resolution status.

Return type:

str

get_searchable_field_statistics(asset_type)

Get the metadata searchable field statistics for a certain asset type.

Parameters:

asset_type – The asset type to receive the searchable field statistics of. To ensure using a valid asset_type, you can import AssetsConstants from scenebox.constants

Returns:

Dictionary with keys of the searchable statistics, and values of the min/max/count/defaults for that statistic.

Return type:

dict

get_metadata(id, asset_type, with_session_metadata=False)

Get the metadata of an asset.

Parameters:
  • id – The ID of the asset to get metadata from.

  • asset_type – The asset type of the asset to get metadata from. To ensure using a valid asset_type, you can import AssetsConstants from scenebox.constants

  • with_session_metadata – If True and the asset belongs to a session, also return session entities

Returns:

The asset metadata.

Return type:

dict

compress_images(ids, desired_shape=None, thumbnail_tag=None, use_preset=None)

Compress a list of images.

Either specify desired_shape or use_preset to compress an image.

Parameters:
  • ids – The IDs of the images to compress.

  • desired_shape – Specifies the desired output shape. Not used when use_preset is set.

  • thumbnail_tag – Tag of the thumbnail to be created. Not used when use_preset is set.

  • use_preset – Use a preset configuration for desired_shape & thumbnail tag. Must be included inside the config.

Returns:

The ID of the job that carries out the image compression.

Return type:

str

compress_videos(ids, desired_shape=None, thumbnail_tag=None, use_preset=None, wait_for_completion=False)

Compress a list of videos.

Either specify desired_shape or use_preset to compress a video.

Parameters:
  • ids – The IDs of the videos to compress.

  • desired_shape – Specifies the desired output shape. Not used when use_preset is set.

  • thumbnail_tag – Tag of the thumbnail to be created. Not used when use_preset is set.

  • use_preset – Use a preset configuration for desired_shape & thumbnail tag. Must be included inside the config.

  • wait_for_completion – should wait for completion?

Returns:

job id of the compression

Return type:

str

compress_lidars(ids, skip_factors=None, gzip_compress=True)

Compress a list of videos.

Either specify desired_shape or use_preset to compress a lidar.

Parameters:
  • ids – The IDs of the LIDARs to compress.

  • skip_factors – Determines how many LIDAR points to skip. Each point listed creates a new thumbnail. Defaults to [1, 10, 100].

  • gzip_compress – Default True. Store a gzip compressed version of the lidars. Allows quicker loading on the webapp.

Returns:

The job ID of the Job that carries out the LIDAR compression.

Return type:

dict

concatenate_videos(ids, output_video_id, video_metadata, job_id=None)

Concatenate a list of videos into a single video.

Parameters:
  • ids – The IDs of the videos to concatenate.

  • output_video_id – The ID of the concatenated video produced.

  • video_metadata – Metadata of the concatenated video produced.

  • job_id – If provided, creates a job with the given job_id. Otherwise, automatically generates a job ID.

Returns:

The job ID of the Job that carries out the video concatenation.

Return type:

dict

add_annotation(annotation, update_asset=True, buffered_write=False, replace_sets=False, add_to_cache=False, annotation_to_objects=False, remove_asset_repeated_annotations=False)

Add a single annotation.

Add an annotation using a class from the scenebox.models.annotation module.

Parameters:
  • annotation – The formatted annotation to ingest.

  • update_asset – If True, updates the metadata of the previously uploaded asset associated with the annotation

  • remove_asset_repeated_annotations – If update_asset is True and remove_asset_repeated_annotations is True, removes repeated annotation listings in the asset’s metadata.

  • buffered_write – If True, ingests annotation in a buffered fashion.

  • replace_sets – If True and update_asset is False, appends to existing annotation metadata rather than replacing it. If False and update_asset is True, replaces existing annotation metadata with the inputted annotation metadata.

  • add_to_cache – If True, corresponding bytes will be added to cache for quick access.

  • annotation_to_objects – If True, runs the annotations_to_objects async task after adding the annotations.

cleanup_assets_annotation_meta(asset_ids, asset_type)

Cleanup annotation lists by removing repeated and non-existent entries in an asset’s metadata. asset_ids:

List of asset ids to clean-up.

asset_type:

The type of asset. To ensure using a valid asset_type, you can import AssetsConstants from scenebox.constants

add_annotations(annotations, buffered_write=True, update_asset=True, threading=True, disable_tqdm=False, replace_sets=False, cleanup_annotation_masks=False, remove_asset_repeated_annotations=False, session_uid=None, annotations_to_objects=False, add_to_cache=False, wait_for_annotations_to_objects_completion=False)

Add several annotations at once.

Parameters:
  • annotations – The annotations to ingest.

  • buffered_write – If True, ingests annotations in a buffered fashion.

  • update_asset – If True, updates the metadata of the asset associated with each of the items in annotations. Otherwise, does not update the metadata of the associated asset.

  • raw_annotations – A list of the raw annotation files to ingest. Gets uploaded as raw files.

  • threading – If True, uses multithreading to speed up annotations ingestion. Otherwise, does not use multithreading.

  • disable_tqdm – If False, uses a progressbar wrapper that calculates and shows the progress of the threading process. Otherwise, disables/silences the tqdm progressbar wrapper.

  • replace_sets – If True and update_asset is False, appends to existing annotation metadata rather than replacing it. If False and update_asset is True, replaces existing annotation metadata with the inputted annotation metadata.

  • cleanup_annotation_masks – If True, runs the cleanup_annotation_masks async task after adding all annotations.

  • remove_asset_repeated_annotations – If update_asset is True and cleanup_asset_meta is True, removes repeated annotation listings in the asset’s metadata.

  • session_uid – If provided and annotations_to_objects is set to True, objects entities are added to the session during the annotations_to_objects async task.

  • annotations_to_objects – If True, runs the annotations_to_objects async task after adding the annotations.

  • wait_for_annotations_to_objects_completion – If True, wait for completion of the annotations_to_objects async task.

  • add_to_cache – If True, corresponding bytes will be added to cache for quick access.

static is_coco_annotation_file_valid(file_path)

Tests if a COCO annotation file is valid.

A valid COCO annotations file is a .json file that only contains keys listed in scenebox.constants.COCOAnnotations.KEYS, and values that are all of type list.

Parameters:

file_path – Location of the COCO annotation file to validate.

Returns:

Returns True if the COCO annotation file is valid (and can be ingested into SceneBox). Otherwise, returns False if the COCO annotation file is invalid.

Return type:

bool

add_coco_annotations_from_folder(file_path, provider, images_set_id, folder_path=None, version=None, annotation_group='ground_truth', session_uid=None, annotations_set_id=None, use_image_id_as_annotation_id=False, preprocesses=None, thumbnailize_at_ingest=False, annotation_to_objects=False, wait_for_completion=False)

Add several images and/or COCO annotations.

Upload several images and/or COCO annotations at once with the same local folder path. This method is best used with local images that are all located in the same folder. For images not located in the same folder, or images that are not located on your local machine, check out self.add_image.

Parameters:
  • file_path – The filepath to the file that contains the coco annotations. Must be in json format.

  • provider – Name of the provider that supplied the annotations.

  • folder_path – The local path to the folder of images to upload. If not provided, no new images are uploaded.

  • version – The version of the model that supplied the annotations.

  • annotation_group – The group that the annotation belongs to. It can be one of ground_truth, model_generated, or other

  • session_uid – The session to associate with the images to upload. If folder_path is not provided, has no effect.

  • images_set_id – The set to add the images to. If folder_path is not provided, has no effect.

  • annotations_set_id – The set to add the COCO annotations to

  • use_image_id_as_annotation_id – If True, appends each annotation’s associated image_id onto the annotation’s ID. Otherwise, makes the annotation IDs None.

  • preprocesses – Specifies which process to treat the image thumbnails with.

  • thumbnailize_at_ingest – If True, create thumbnail at ingestion time. Otherwise, create thumbnails “on the fly”.

  • annotation_to_objects – If True, extract objects from the annotations when adding them.

  • wait_for_completion – If True, wait for the completion of the annotation to object process

delete_annotations(ids=None, search=None)

Delete annotations using a list of ids or a search query, and updates metadata of corresponding assets.

Parameters:
  • ids – A list of annotation IDs to delete.

  • search – Query to locate the data subset of interest. Filters through existing assets according to the dictionary passed.

Returns:

A list of job IDs. One for each job that carries out the deletion of a chunk of annotations.

Return type:

job_ids

remove_annotation_sources(annotation_sources, asset_type, asset_filters=None, wait_for_completion=True)

Delete annotations using a list of ids or a search query, and updates metadata of corresponding assets.

Parameters:
  • annotation_sources – A list of dictionaries. Each dictionary defines an annotation source with the mandatory field: provider and optional fields: version, annotation_type, annotation_group. Can be collected from the get_annotation_sources method.

  • asset_type – Asset media type for annotations provided by all annotation_sources. To ensure using a valid asset_type, you can import AssetsConstants from scenebox.constants

  • asset_filters – Query to locate the data subset of interest. Filters through existing assets according to the dictionary passed.

  • wait_for_completion – If True, polls until job is complete. Otherwise, continues execution and does not raise an error if the job does not complete. Default is True.

Returns:

The job ID of the Job that carries out the deletion of annotations.

Return type:

job_id

add_session(session_name, session_type='generic', set_id=None, resolution=None, session_aux_metadata=None, status='ready', raise_if_session_exists=False)

Add a session.

Parameters:
  • session_name – The name to give to the session.

  • session_type – The session type to give to the session. Choose from scenebox.constants.SessionTypes.

  • set_id – The session set to add the session to.

  • resolution – The resolution at which to sample events on. Measured in seconds. Ensure you choose a small enough resolution that none of your events get aliased to occur at a lower frequency than expected.

  • session_aux_metadata – session auxiliary metadata including any metadata that are not primary on SceneBox and user wants to search sessions with Example:

    {

    “key_1”: 1, “key_2”: 1.0, “key_3”: “1”, “key_4”: [“abc”, “def”], “key_5”: {“a”: 1, “b”: 2}

    }

  • status – Session status. By default, this is READY but can be IN PROGRESS as well.

  • raise_if_session_exists – Raises a SessionError if the name parameter matches an existing set.

Returns:

The session UID of the added session.

Return type:

str

slice_session(session_uid, start_time, end_time, session_name=None)

slice an existing session with start and end time.

Parameters:
  • session_uid – session id to be sliced

  • start_time – start time stamp of the slice

  • end_time – end time stamp of the slice

  • session_name – Optional new name for the silced_session

Returns:

The session UID of the added session.

Return type:

str

update_time_span(session_uid, session_start_time, session_end_time)

Update the start/end times of a session.

Parameters:
  • session_uid – session uid of the session to change the start/end time for.

  • session_start_time – The new session start time.

  • session_end_time – The new session end time.

Returns:

The updated metadata of the session after the timespan update.

Return type:

dict

extend_time_span(session_uid, timestamps)

Extend the start time and end time of a session based on min/max of a list of timestamps.

Parameters:
  • session_uid – session uid of the session to change the start/end time for.

  • timestamps – list of timestamps

Returns:

The updated metadata of the session after the timespan update.

Return type:

dict

add_source_data(session_uid, source_data, sensors, replace=True)

Add source data to session metadata.

Parameters:
  • session_uid – The session UID of the session metadata to update.

  • source_data – The source data to add to the session metadata. Example: [

    {

    “sensor”: “camera_1”, “id”: “id_1”, “type”: “videos”

    }, {

    “sensor”: “camera_2”, “id”: “id_2”, “type”: “videos”

    }

    ]

  • sensors – The sensors to be added to session metadata. Describes what sensors were used to capture the data. Example: [

    {

    “name”: “camera_1”, “type”: “camera”

    }, {

    “name”: “camera_2”, “type”: “camera”

    }

    ]

  • replace – If True, overwrites existing source metadata with the supplied source data. Otherwise, appends the supplied source data to the existing source metadata.

Returns:

The updated metadata of the session after the source data update.

Return type:

dict

add_auxiliary_session_data(session_uid, auxiliary_data, replace=True)

Add auxiliary session data e.g. video data to metadata of a given session.

Parameters:
  • session_uid – The session UID Of the session to add auxiliary data to.

  • auxiliary_data – The auxiliary data to add to the session metadata. Example: [

    {

    “id”: “id_1”, “type”: “concatenated_video”, “tags”: [

    “camera_1”

    ]

    }, {

    “id”: “id_2”, “type”: “concatenated_video”, “tags”: [

    “camera_2”

    ]

    }

    ]

  • replace – If True, overwrites existing auxiliary metadata with the supplied auxiliary data. Otherwise, appends the supplied auxiliary data to the existing auxiliary metadata.

Returns:

The updated metadata of the session after the auxiliary session data update.

Return type:

dict

add_ego_pose(session_uid, ego_pose_data=None, ego_pose_path=None, title=None)

Add an ego pose file to a session. An ego pose file has a series of frames in which a list of timestamped objects with their ego position in “lat” and “lon” is given. Example:

[
{

“frame”:0, “timestamp”:”2021-08-24T21:15:15.574000+00:00”, “objects”:{

“0”:{

“lat”:33.6, “lon”:-0.15

}, “1”:{

“lat”:5.6, “lon”:-3.45

}, “2”:{

“lat”:23.25, “lon”:-3.85

}, “3”:{

“lat”:83.0, “lon”:-5.9

}, “4”:{

“lat”:198.55, “lon”:12.15

}

}

}, {

“frame”:1, “timestamp”:”2021-08-24T21:15:15.599000+00:00”, “objects”:{

“5”:{

“lat”:33.55, “lon”:-0.15

}, “6”:{

“lat”:5.7, “lon”:-3.45

}, “7”:{

“lat”:23.25, “lon”:-3.85

}, “8”:{

“lat”:82.45, “lon”:-5.9

}, “9”:{

“lat”:197.8, “lon”:12.15

}

}

}

]

session_uid:

The session UID Of the session to add auxiliary ego pose of the objects.

ego_pose_data:

dictionary for ego pose of the objects

ego_pose_path:

path to a json file containing the json file for ego pose of the objects.

title:

title of the ego pose attached to the session.

add_enrichments_configs(session_uid, enrichments_configs, replace=True)

Add enrichment configurations to session metadata.

Parameters:
  • session_uid – The session UID of the session to add the enrichment configurations to.

  • enrichments_configs – The enrichments configuration to add to the session metadata. Example: [

    {

    “input_event”: “geo_locations”, “type”: “location”, “configuration”: {

    “influence_type”: “interval”, “influence_radius_in_seconds”: 0.5

    }

    }, {

    “input_event”: “geo_locations”, “type”: “visibility”, “configuration”: {

    “influence_type”: “interval”, “influence_radius_in_seconds”: 2.0

    }

    }, {

    “input_event”: “geo_locations”, “type”: “weather”, “configuration”: {

    “influence_type”: “interval”, “influence_radius_in_seconds”: 2.0

    }

    }

    ]

  • replace – If True, overwrites existing enrichments config metadata with the supplied enrichments config. Otherwise, appends the supplied enrichments config to the existing enrichments config metadata.

Returns:

The updated metadata of the session after adding the enrichments configs.

Return type:

dict

add_geolocation_entity(latitude, longitude, timestamp, session_uid, buffered=False)

Add a geolocation entity to a session.

Parameters:
  • latitude – The latitude of the geolocation.

  • longitude – The longitude of the geolocation.

  • timestamp – The timestamp of the geolocation.

  • session_uid – The session uid of the session to add the geolocation entity to.

  • buffered – Buffer write

Returns:

A confirmation that the Job to add the entity was acknowledged.

Return type:

dict

add_geolocation_entities(latitudes, longitudes, timestamps, session_uid, buffered=False)

Add geolocation entities to a session.

Parameters:
  • latitudes – A list of latitudes of the geolocations.

  • longitudes – A list of longitudes of the geolocations.

  • timestamps – A list of timestamps of the geolocations.

  • session_uid – The session uid of the session to add the geolocation entities to.

  • buffered – Buffer write

Returns:

A list of dictionaries containing confirmations that the Job to add the entity chunks were acknowledged.

Return type:

List

add_entity(entity_dict, urgent=False, buffered=False)

Add an entity to the specified session.

Parameters:
  • entity_dict – The entity data to add. Typically, has the following form: {

    “session”: session_uid, “start_time”: datetime_or_str_to_iso_utc(start_time), “end_time”: datetime_or_str_to_iso_utc(end_time), “timestamp”: datetime_or_str_to_iso_utc(start_time), “entity_id”: entity_id, “entity_name”: event_name, “entity_value”: event_value, “entity_type”: Choose from scenebox.constants.EntityTypes, “influence”: Choose from scenebox.constants.EntityInfluenceTypes “extend_session”: bool whether to extend a session to cover the influence interval

    }

  • urgent – If True, entity is ingested immediately. However, a manual resolution is needed afterwards to make the entity searchable.

  • buffered – If true, ignores the urgent and uses a built-in buffer for writing.

Returns:

A confirmation that the Job to add the entity was acknowledged.

Return type:

dict

add_entities(entity_dicts, urgent=False, buffered=False)

Add a list of entities to the specified sessions.

Parameters:
  • entity_dicts – The entity data to add. Typically, each list item has the following form: {

    “session”: session_uid, “start_time”: datetime_or_str_to_iso_utc(start_time), “end_time”: datetime_or_str_to_iso_utc(end_time), “timestamp”: datetime_or_str_to_iso_utc(start_time), “entity_id”: entity_id, “entity_name”: event_name, “entity_value”: event_value, “entity_type”: Choose from scenebox.constants.EntityTypes, “influence”: Choose from scenebox.constants.EntityInfluenceTypes “extend_session”: bool whether to extend a session to cover the influence interval

    }

  • urgent – If True, entity is ingested immediately. However, a manual resolution is needed afterwards to make the entity searchable.

  • buffered – if True, the entities is written to the buffer and “urgent” is ignored. user should call flush_entities_buffer eventuially to write all the data.

Returns:

A confirmation that the Job to add the entities was acknowledged.

Return type:

dict

flush_entities_buffer()

write all of the entities in the buffer. If scene engine is exited without calling this, all of the unflushed entities are discarded.

resolve_session(session_uid, resolution=None, wait_for_completion=False)

Resolve a session.

Project session events onto a single timeline. Events are sampled at the given resolution. A smaller resolution is recommended if events rapidly change in value.

Parameters:
  • session_uid – The session UID of the session to resolve.

  • resolution – The resolution at which to sample session events. Measured in seconds.

  • wait_for_completion – If True, polls until job is complete. Otherwise, continues execution and does not raise an error if the job does not complete.

Returns:

The ID of the Job that attempts to resolve the session.

Return type:

dict

add_event_interval(event_name, event_value, start_time, end_time, session_uid, urgent=False, buffered=False)

Add an event interval to a session.

Adds a “state_set” entity to a specific session’s timespan.

Parameters:
  • event_name – The name of the event interval to add to the session.

  • event_value – The value(s) of the event over the timespan.

  • start_time – The start time of the event.

  • end_time – The end time of the event.

  • session_uid – The session UID of the session to add the event to.

  • urgent – If True, entity is ingested immediately. However, a manual resolution is needed afterwards to make the entity searchable.

  • buffered – buffered write

Returns:

The ID of the added entity.

Return type:

str

add_comments(comments, start_time, end_time, session_uid, wait_for_completion=False)

Add a comment to a time segment of a session.

Parameters:
  • comments – The comment(s) to add to the given timespan.

  • start_time – The start time of the timespan to add the comment to.

  • end_time – The end time of the timespan to add the comment to.

  • session_uid – The session UID of the session to add the comment to.

  • wait_for_completion – If True, polls until job is complete. Otherwise, continues execution and does not raise an error if the job does not complete.

Returns:

The ID of the Job that attempts to add the comment.

Return type:

dict

add_event_intervals(event_name, event_values, start_times, end_times, session_uid, urgent=False, extend_session=True, epsilon=0, buffered=False)

Add a list of several event intervals to a session.

Adds several “state_set” entities to a specific session’s timespan.

Parameters:
  • event_name – The name of the event interval to add to the session.

  • event_values – The value(s) of the event over the timespan.

  • start_times – The start time of each event.

  • end_times – The end time of each event.

  • session_uid – The session UID of the session to add the events to.

  • urgent – If True, entity is ingested immediately. However, a manual resolution is needed afterwards to make the entity searchable.

  • extend_session – should extend the session to include this event (default to true)

  • epsilon – Constant to increase the time-delta of entity start time and endtime. Measured in seconds.

  • buffered – buffered write

Returns:

A list of the added entity IDs

Return type:

List[str]

add_scalar_intervals(measurement_name, measurement_values, start_times, end_times, session_uid, epsilon=0.001)

Add a timeseries with interval values to a session.

Parameters:
  • measurement_name – The name to assign to the measurements.

  • measurement_values – The numerical values of each interval. Each interval can only have one value.

  • start_times – The start times of each interval.

  • end_times – The end times of each interval.

  • session_uid – The UID of the session to add the timeseries to.

  • epsilon – The amount of time to add to start times and subtract off end times. Prevents intervals from having undesired overlap.

Returns:

The IDs of the created entities.

Return type:

List[str]

add_scalar_measurements(measurement_name, measurement_values, timestamps, session_uid, buffered=False)

Add a timeseries with point values (“scalars”) to a session.

Parameters:
  • measurement_name – The name to assign to the measurements.

  • measurement_values – The numerical values of each point. Each point can only have one value.

  • timestamps – The timestamps at which each measurement occurred.

  • session_uid – The UID of the session to add the timeseries to.

  • buffered – If True, writes events in a buffered fashion.

Returns:

The IDs of the created entities.

Return type:

List[str]

add_point_events(measurement_name, measurement_values, timestamps, session_uid, buffered=False)

Add a list of several point events to a session.

Parameters:
  • measurement_name – The name to assign to the measurements.

  • measurement_values – The state of the event at each timestamp.

  • timestamps – A list of timestamps at which the events occurred.

  • session_uid – The UID of the session to add the events to.

  • buffered – If True, writes events in a buffered fashion.

Returns:

A list of the added entity IDs.

Return type:

List[str]

add_timeseries_csv(session_uid, csv_filepath, df_labels=None)

Add a timeseries from a CSV file.

Add a measurement DataFrame to a session as scalar entities from a CSV. Creates a measurement_df Pandas DataFrame from the inputted CSV file, then passes this dataframe to self.add_df.

Parameters:
  • session_uid – The session to add the timeseries to.

  • csv_filepath – The filepath of the CSV to turn into a timeseries.

  • df_labels – List of the CSV column names to use.

add_df(measurement_df, session_uid, timestamps=None)

Add a measurement DataFrame to a session as scalar entities.

Add several measurements across time in a session by populating the measurement_df Pandas DataFrame argument. Can add an arbitrary number of named measurements, either numeric or non-numeric. Timestamps must either be specified in the timestamps argument, or under a column named “timestamps” inside measurement_df.

Parameters:
  • measurement_df – DataFrame holding the measurement(s) of interest. Add an arbitrary number of named columns to represent different measurement types. If a column named “timestamp” specifying the timestamps of the measurements does not exist, must be specified under the timestamps method argument.

    If latitude/longitude columns are included in under the names “lat” and “lon” respectively, will automatically be added as geolocation entities.

  • session_uid – The session UID to add the measurement(s) to.

  • timestamps – A list of timestamps corresponding to the data measurements in ``measurement_df`

entity_summary(session_uid, entity_names=None, start_time=None, end_time=None)

Get the entity summary for a session.

Return type:

dict

delete_session(session_uid, delete_assets_contents=True, wait_for_completion=False)

Delete an existing session.

Optionally delete the assets inside the session, as well.

Parameters:
  • session_uid – The UID of the session to delete.

  • delete_assets_contents – If True, deletes the assets contained inside the specified session. Otherwise, does not delete the assets in the session.

  • wait_for_completion – If True, polls until job is complete. Otherwise, continues execution and does not raise an error if the job does not complete.

  • Returns

search_assets(asset_type, search=None, size=50, offset=0, sort_field=None, sort_order=None, scan=False)

Retrieve asset IDs with a search query.

Returns the top size matching hits. If a return of more than 10000 hits is desired, please use AssetManagerClient.search_assets_large().

Parameters:
  • asset_type – Asset type to filter for in the asset search. To ensure using a valid asset_type, you can import AssetsConstants from scenebox.constants

  • search – Query to locate the data subset of interest. Filters through existing assets according to the dictionary passed.

  • size – Specifies the Elasticsearch search size. The maximum number of hits to return. Has no effect if scan is False.

  • offset – Specifies the Elasticsearch search offset. The number of hits to skip. Has no effect if scan is False.

  • sort_field – Sorts for the specified name.

  • sort_order – Specifies the Elasticsearch string sorting order.

  • scan – If True, uses the Elasticsearch scan capability. Otherwise, uses the Elasticsearch search API.

Returns:

A list of the IDs of the assets fulfilling the search query.

Return type:

List[str]

search_meta(asset_type, query=None, size=50, offset=0, sort_field=None, sort_order=None, scan=False, compress=False)

Retrieve asset metadata with a search query.

Returns the top size matching hits. If a return of more than 10000 hits is desired, please use AssetManagerClient.search_meta_large().

Parameters:
  • asset_type – Asset type to filter for in the asset metadata search. To ensure using a valid asset_type, you can import AssetsConstants from scenebox.constants

  • query – Query to locate the data subset of interest. Filters through existing assets according to the dictionary passed.

  • size – Specifies the Elasticsearch search size. The maximum number of hits to return. Has no effect if scan is False.

  • offset – Specifies the Elasticsearch search offset. The number of hits to skip. Has no effect if scan is False.

  • sort_field – Sorts for the specified name.

  • sort_order – Specifies the Elasticsearch string sorting order.

  • scan – If True, uses the Elasticsearch scan capability. Otherwise, uses the Elasticsearch search API.

  • compress – Boolean. If set to True, a gzip compressed list of metadata is returned. Typically used in cases where the metadata returned is over 20MB.

Returns:

A list of the metadata of the assets fulfilling the search query.

Return type:

List[Dict]

count(asset_type, search=None)

Retrieve summary of asset metadata with a search query.

Parameters:
  • asset_type – Asset type to collect metadata summary for. To ensure using a valid asset_type, you can import AssetsConstants from scenebox.constants

  • search – Query to locate the data subset of interest. Filters through existing assets according to the dictionary passed.

Return type:

int

summary_meta(asset_type, summary_request)

Retrieve summary of asset metadata with a search query.

Parameters:
  • asset_type – Asset type to collect metadata summary for. To ensure using a valid asset_type, you can import AssetsConstants from scenebox.constants

  • summary_request

    Includes:
    • search: (Dict) to locate the data subset of interest.

    • dimensions: (List[str]) to collect statistical summary of the desired dimensions

    • nested_dimensions: (List[str]) to collect statistical summary of the desired nested dimensions

    Example:
    {
    “search”: {
    “filters”: [
    {

    “field”: “objects”, “values”: [

    “ball.camera_0.tester_manual_point”

    ], “filter_out”: false

    }

    ]

    }, “nested_dimensions”: [

    “annotations_meta.provider”

    ]

    }

Returns:

Example: {

”total”: 3, “aggregations”: [

{

“dimension”: “annotations_meta.provider”, “buckets”: [

{

“key”: “COCO_bounding_box”, “doc_count”: 1

}, {

”key”: “COCO_polygon”, “doc_count”: 2

}

]

}

]

}

Return type:

Dict

add_embeddings(embeddings, similarity_ingest=True, create_umap=True, wait_for_completion=False, add_to_cache=False, umap_params=None)

Add existing embeddings to an asset.

Parameters:
  • embeddings – List of embeddings objects that are created from Scenebox Embedding class.

  • similarity_ingest – If True, enables similar image/object search by performing bulk indexing. Otherwise, has no effect.

  • create_umap – If True, the embeddings are going to be added to a new umap

  • wait_for_completion – If True, polls until job is complete. Otherwise, continues execution and does not raise an error if the job does not complete.

  • add_to_cache – If True, corresponding bytes will be added to cache for quick access.

  • umap_params – umap parameters including: n_neighbors, min_dist and train_size. Default is {

    n_neighbors: 20 min_dist: 0.3 train_size: 10000

    }

Returns:

The IDs of the successfully added embeddings

Return type:

List[str]

add_image(image_path=None, id=None, image_url=None, image_uri=None, image_bytes=None, sensor=None, timestamp=None, session_uid=None, set_id=None, annotations=None, preprocesses=None, aux_metadata=None, geo_field=None, shape_group_field=None, nested_fields=None, enrichments=None, add_to_session=False, thumbnailize_at_ingest=False, buffered_write=False, add_provider_to_labels=True, overwrite=True, add_to_cache=False)

Upload an image onto SceneBox.

Upload an image with a local file path, URL, URI, or image bytes. This method is best used with singular images, or images that are not all located in the same folder. For several images all located in the same folder, check out self.add_images_from_folder.

Parameters:
  • image_path – The local path to the image to upload. If not None, image_url, image_uri, and image_bytes should all be None.

  • id – A unique image ID. A common choice is the image filename.

  • image_url – The URL of the image to upload. Images must be publicly available. If not None, image_url, image_uri, and image_path should all be None.

  • image_uri – The URI of the image to upload. Can be from a private source. If not None, image_url, image_path, and image_bytes should all be None.

  • image_bytes – The image bytes to upload. If not None, image_path, image_url, and image_uri should all be None.

  • sensor – The sensor associated with the image.

  • timestamp – The time at which the image was taken.

  • session_uid – The session associated with the image.

  • set_id – The set (str) or sets (List[str]) to add the image to.

  • annotations – Annotations associated with the image. Each item in the passed list must be of a class from scenebox.models.annotation.

  • preprocesses – Specifies which process to treat the image thumbnails with.

  • aux_metadata – Auxiliary metadata associated with the image (partitioned from primary metadata)

  • geo_field – Geolocation metadata associated with the image.

  • shape_group_field – Shape group metadata associated with the image (example: UMAP).

  • nested_fields – nested fields (example: “[annotations_meta]”)

  • enrichments – The types of enrichments to add to the image.

  • thumbnailize_at_ingest – If True, create thumbnail at ingestion time. Otherwise, create thumbnails “on the fly”.

  • buffered_write – If True, writes images in a buffered fashion.

  • add_to_session – If True and session_uid is not None, add to the existing session with the passed session_uid.

  • add_provider_to_labels – If True the labels of the annotations in aux.label will included the provider as well. If False only the labels will be ingested

  • overwrite – If True and id is not None, updates the metadata/annotations/etc. of a previously uploaded image.

  • add_to_cache – If True, corresponding bytes will be added to cache for quick access.

Returns:

The id of the added image.

Return type:

str

Raises:
  • SetsErrorInvalidArgumentsError: – If more than one of image_path, image_url, image_uri, and image_bytes is not None

  • ValueError: – If image_path, image_bytes, image_uri, and image_url are all None

add_images_from_folder(folder_path, set_id, session_uid=None, filename_image_id_map=None, preprocesses=None, thumbnailize_at_ingest=False)

Upload images from a single local folder path onto SceneBox.

Upload several images at once with the same local folder path. This method is best used with local images that are all located in the same folder. For images not located in the same folder, or images that are not located on your local machine, check out self.add_image.

Parameters:
  • folder_path – The local path to the folder of images to upload.

  • session_uid – The session associated with the image.

  • set_id – The set to add the images to.

  • filename_image_id_map – Provides a mapping from the image filenames to the desired image IDs. If not specified, Image IDs are automatically assigned to a randomized, unique string.

  • preprocesses – Specifies which process to images thumbnails with.

  • thumbnailize_at_ingest – If True, create thumbnail at ingestion time. Otherwise, create thumbnails “on the fly”.

Raises:

ValueError: – “If image_path, image_bytes, image_uri, and image_url are all None”

add_images(images, set_id, geo_field=None, shape_group_field=None, nested_fields=None, preprocesses=None, add_to_session=False, thumbnailize_at_ingest=False, add_provider_to_labels=True, add_to_cache=False, overwrite=True)

Upload multiple images onto SceneBox.

Upload an image with a local file path, URL, URI, or image bytes. This method is best used with singular images, or images that are not all located in the same folder. For several images all located in the same folder, check out self.add_images_from_folder.

Parameters:
  • images – A list of objects of the type Image defined in models.Image. For Image objects that exist, the metadata is overwritten with the information provided in this call. For updating metadata, refer to asset_manager_client.update_metadata().

  • set_id – The set (str) or sets (List[str]) to add the images to.

  • geo_field – Geolocation metadata associated with the image.

  • shape_group_field – Shape group metadata associated with the image (example: UMAP).

  • nested_fields – nested fields (example: “[annotations_meta]”)

  • preprocesses – Specifies which process to treat the image thumbnails with.

  • thumbnailize_at_ingest – If True, create thumbnail at ingestion time. Otherwise, create thumbnails “on the fly”.

  • add_to_session – If True and session_uid is not None, add to the existing session with the passed session_uid.

  • add_provider_to_labels – If True the labels of the annotations in aux.label will included the provider as well. If False only the labels will be ingested

  • add_to_cache – If True, corresponding bytes will be added to cache for quick access.

  • overwrite – If False and asset exists on SceneBox, adds the previously uploaded images to the given set_id.

Returns:

List of ids of the added images.

Return type:

List[str]

Raises:
  • SetsErrorInvalidArgumentsError: – If more than one of image_path, image_url, image_uri, and image_bytes is not None

  • ValueError: – If image_path, image_bytes, image_uri, and image_url are all None

add_video(set_id, video_path=None, video_url=None, video_uri=None, sensor=None, start_timestamp=None, session_uid=None, id=None, annotations=None, aux_metadata=None, enrichments=None, tags=None, compress_video=True, buffered_write=False, add_to_session=False, create_session=False)

Upload a video onto SceneBox.

Upload a video with a local file path, URL, or URI.

Parameters:
  • video_path – The local path to the video to upload. If not None, video_uri and video_url should both be None.

  • video_url – The URL of the video to upload. Video must be publicly available. If not None, video_path and video_uri should both be None.

  • video_uri – The URI of the video to upload. Can be from a private source. If not None, video_path, and video_url, should both be None.

  • sensor – The sensor associated with the image.

  • start_timestamp – The time at which the video recording began.

  • session_uid – The session associated with the video.

  • id – A unique video ID. A common choice is the video filename.

  • set_id – The set to add the video to.

  • annotations – Annotations associated with the Video. Each item in the passed list must be of a class from scenebox.models.annotation.

  • aux_metadata – Auxiliary metadata associated with the image (partitioned from primary metadata)

  • enrichments – The types of enrichments to add to the video.

  • tags – Labels associated with the video. Allows for easy video search.

  • compress_video – If true, register compressed video thumbnails.

  • buffered_write – If True, writes videos in a buffered fashion.

  • add_to_session – If True and session_uid is not None, add to the existing session with the passed session_uid.

  • create_session – If True and session_uid is None, create a single video session from the video with the video naming and the video start time and duration. Session resolution would be default (1.0 second) and video will be added as an aux_metadata. In this case, if the sensor name is not specified, it will be “main”

Returns:

The ID of the added video.

Return type:

str

Raises:

ValueError: – If video_path, video_url, and video_uri are all None

add_videos(videos, set_id, add_to_session=False, compress_videos=True, add_to_cache=False, overwrite=True, wait_for_compression_completion=False)

Upload multiple videos onto SceneBox.

Parameters:
  • videos – A list of objects of the type Video defined in models.Video. For Video objects that exist, the metadata is overwritten with the information provided in this call. For updating metadata, refer to asset_manager_client.update_metadata().

  • set_id – The set (str) or sets (List[str]) to add the videos to.

  • add_to_session – If True and session_uid is not None, add to the existing session with the passed session_uid.

  • compress_videos – If true, register compressed video thumbnails.

  • add_to_cache – If True, corresponding bytes will be added to cache for quick access.

  • overwrite – If False and asset exists on SceneBox, adds the previously uploaded videos to the given set_id.

  • wait_for_compression_completion – If True wait for compression to be completed

Returns:

List of ids of the added videos.

Return type:

List[str]

Raises:

SetsErrorInvalidArgumentsError: – If more than one of video path, url, uri, and bytes is not None

get_video_frame_thumbnail(video_id, frame, thumbnail_tag)

Get thumbnail png for requested frame fram a video.

Parameters:
  • video_id – ID of the video to fetch the frame from.

  • frame – The number ‘n’ for the n-th frame from the beginning.

  • thumbnail_tag – Tag of the thumbnail to be created.

Returns:

URL for the requested thumbnail

Return type:

str

extract_frame_thumbnails(ids, frames, thumbnail_tags=None, wait_for_completion=False)

Extract thumbnails for a list of frames in a video.

Parameters:
  • ids – IDs of the videos to extract frame thumbnails from.

  • frames – A list of frame numbers starting from the beginning of each video.

  • thumbnail_tags – List of types of thumbnails tags to extract.

Returns:

The job ID of the Job that running the operation.

Return type:

str

add_lidar(set_id, lidar_path=None, lidar_url=None, lidar_uri=None, lidar_bytes=None, sensor=None, timestamp=None, session_uid=None, id=None, format='pcd', binary_format=None, num_fields=None, enrichments=None, buffered_write=False, add_to_session=False, add_to_cache=False, aux_metadata=None)

Upload a LIDAR onto SceneBox.

Upload a LIDAR with a local file path, URL, or URI.

Parameters:
  • lidar_path – The local path to the LIDAR to upload. If not None, lidar_uri and lidar_url should both be None.

  • lidar_url – The URL of the LIDAR to upload. LIDAR must be publicly available. If not None, lidar_path and lidar_uri should both be None.

  • lidar_uri – The URI of the LIDAR to upload. Can be from a private source. If not None, lidar_path, and lidar_url, should both be None.

  • sensor – The sensor associated with the LIDAR.

  • timestamp – The time at which the LIDAR was taken.

  • session_uid – The session associated with the LIDAR.

  • id – A unique LIDAR ID. A common choice is the LIDAR filename.

  • format – The format in which the LIDAR was embedded. pcd|numpy|binary. Default is pcd. PCD files are expected to be valid. Support for binary files is experimental.

  • binary_format – Experimental. For binary data, a decoder to decode each packet to x,y,z,intensity,epoch. Format should be compatible with struct (https://docs.python.org/3/library/struct.html) library.

  • num_fields – The number of fields exists in LIDAR file to reshape numpy array for displaying. Required for numpy formats

  • set_id – The set to add the LIDAR to.

  • enrichments – The types of enrichments to add to the LIDAR.

  • buffered_write – If True, writes images in a buffered fashion.

  • add_to_session – If True and session_uid is not None, add to the existing session with the passed session_uid.

  • add_to_cache – If True, corresponding bytes will be added to cache for quick access.

  • aux_metadata – Auxiliary metadata associated with the image (partitioned from primary metadata)

Returns:

The ID of the added LIDAR.

Return type:

str

Raises:

ValueError: – If lidar_path, lidar_url, and lidar_uri are all None.

add_lidars(lidars, set_id, add_to_session=False, add_to_cache=False, overwrite=True)

Upload multiple videos onto SceneBox.

Parameters:
  • lidars – A list of objects of the type Lidar defined in models.Lidar. For Lidar objects that exist, the metadata is overwritten with the information provided in this call. For updating metadata, refer to asset_manager_client.update_metadata().

  • set_id – The set (str) or sets (List[str]) to add the lidars to.

  • add_to_session – If True and session_uid is not None, add to the existing session with the passed session_uid.

  • add_to_cache – If True, corresponding bytes will be added to cache for quick access.

  • overwrite – If False and asset exists on SceneBox, adds the previously uploaded lidars to the given set_id.

Returns:

List of ids of the added lidars.

Return type:

List[str]

Raises:

SetsErrorInvalidArgumentsError: – If more than one of lidar path, url, uri, and bytes is not None

enable_webapp_features(features)

Enable a list of features on SceneBox WebApp.

Parameters:

features – A list of features to enable on SceneBox WebApp. For example: [“UPLOAD_RECORDING”]

Raises:

AssetError: – If multiple org_configs are found. If org config exists but the name does not have a supported format.

disable_webapp_features(features)

Disable a list of features on SceneBox WebApp.

Parameters:

features – A list of features to disable on SceneBox WebApp. For example: [“UPLOAD_RECORDING”]

Raises:

AssetError: – If multiple org_configs are found. If org config exists but the name does not have a supported format.

index_s3_images_batch(bucket, folder, thumbnailize_at_ingest=True, overwrite=True, wait_for_completion=False, extensions=['jpeg', 'jpg', 'png', 'tif', 'tiff'], max_wait_sec=None)

Index s3 images.

Ingest images in a given s3 bucket and folder.

Parameters:
  • bucket – The bucket name containing images for example “my_bucket”.

  • folder – The folder containing images for example “folder” or “folder/sub_folder”.

  • thumbnailize_at_ingest – If True, create thumbnail at ingestion time. Otherwise, create thumbnails “on the fly”.

  • overwrite – If True, re-index the image which will overwrite an existing image with the same id.

  • wait_for_completion – If True, polls until job is complete. Otherwise, continues execution and does not raise an error if the job does not complete.

  • extensions – Extensions of images to ingest. Pass “*” to ingest all objects with/without any extension.

  • max_wait_sec – If a value is given, the wait for job completion will be for that many seconds. If None (default), and wait_for_completion is True, there will be no limit on the wait for job completion.

Returns:

The associated job id

Return type:

str

index_recording(path, recording_name=None, session_set_id=None, session_aux_metadata=None, wait_for_completion=False, max_wait_sec=None)

Index Recording.

Ingests recording data into scenebox including avi videos, etc.

Parameters:
  • path – The path containing all including the bucket e.g. bucket/a/b/recording_name or bucket/recording_name

  • recording_name – The name of the recording to ingest. This will be used as session name. If not given, will get that info from path

  • session_set_id – The id of the existing session set id to add the new session to.

  • session_aux_metadata – The auxiliary metadata to be added to session’s metadata

  • wait_for_completion – If True, polls until job is complete. Otherwise, continues execution and does not raise an error if the job does not complete.

  • max_wait_sec – If a value is given, the wait for job completion will be for that many seconds. If None (default), and wait_for_completion is True, there will be no limit on the wait for job completion.

Returns:

The associated job id

Return type:

str

index_rosbag(sets_prefix, rosbag_uri=None, rosbag_path=None, session_name=None, session_uid=None, session_set_id=None, session_aux_metadata=None, deep_analysis=False, models=None, sensors=None, topics=None, session_resolution=1.0, image_skip_factor=10, indexing_parameters=None, wait_for_completion=False, max_wait_sec=None)

Index Ros1 files.

Ingests ros1 data into scenebox, then extracts all available images, videos, and other asset types.

Parameters:
  • sets_prefix – The prefix to append on all created set names.

  • rosbag_uri – The uri location of the folder containing the rosbag file If not None, rosbag_path should be None.

  • rosbag_path – The local path of the rosbag If not None, rosbag_uri should be None.

  • session_name – The name to give to the session or added previously to SceneBox which is associated with the indexed rosbag.

  • session_uid – The uid of the session added previously to SceneBox and will be associated with the indexed rosbag. If not provided, the session will be added.

  • session_set_id – The id of the existing session set id to add the new rosbag session to.

  • session_aux_metadata – The auxiliary metadata to be added to session’s metadata

  • deep_analysis – Enables enrichments such as object extraction, image/object embeddings, and UMAP visualizaion.

  • models – A list of the models to use to enrich the rosbag data. Choose from the models listed in scenebox.constants.AssetsConstants.MLModelConstants.

  • sensors – A list of the sensors (listed in image metadata) to create videos from.

  • topics – A list topics to ingest. If no topics passed, all topics will be ingested.

  • session_resolution – resolution of the ingested session in seconds

  • image_skip_factor – skip factor for image messages

  • indexing_parameters – indexing parameters for the rosbag ingestion

  • wait_for_completion – If True, polls until job is complete. Otherwise, continues execution and does not raise an error if the job does not complete.

  • max_wait_sec – If a value is given, the wait for job completion will be for that many seconds. If None (default), and wait_for_completion is True, there will be no limit on the wait for job completion.

Returns:

The associated job id

Return type:

str

index_rosbag2(sets_prefix, relative_file_paths, folder_uri=None, folder_path=None, session_name=None, session_set_id=None, deep_analysis=False, models=None, sensors=None, topics=None, indexing_parameters=None, session_resolution=1.0, wait_for_completion=False)

Index Ros2 files.

Ingests ros2 data into scenebox, then extracts all available images, videos, and other asset types.

Parameters:
  • sets_prefix – The prefix to append on all created set names.

  • relative_file_paths – The relative file paths of rosbag files to ingest (.yaml/db3 files). Relative according to specified

  • folder_uri – The uri location of the folder containing the rosbag db3 file and metadata yaml file. If not None, folder_path should be None.

  • folder_path – The local folder path of the folder containing the rosbag db3 file and metadata yaml file. If not None, folder_uri should be None.

  • session_name – The name to give to the session associated with the indexed ros2.

  • session_set_id – The id of the existing session set id to add the new ros2 session to.

  • deep_analysis – Enables enrichments such as object extraction, image/object embeddings, and UMAP visualizaion.

  • models – A list of the models to use to enrich the ros2 data. Choose from the models listed in scenebox.constants.AssetsConstants.MLModelConstants.

  • sensors – A list of the sensors (listed in image metadata) to create videos from.

  • topics – A list topics to ingest. If no topics passed, all topics will be ingested.

  • indexing_parameters – indexing parameters for the rosbag ingestion

  • session_resolution – resolution of the ingested session in seconds

  • wait_for_completion – If True, polls until job is complete. Otherwise, continues execution and does not raise an error if the job does not complete.

Returns:

A dictionary containing “job_id” as the key and the associated job id as the value.

Return type:

dict

add_rosbag2(relative_file_paths, folder_uri=None, folder_path=None, session_uid=None, set_id=None)

Upload Rosbag2 onto SceneBox.

Upload Rosbag2 with a local file path, URL, or URI. Can only process ros2.

Parameters:
  • relative_file_paths – The relative file paths of rosbag files to ingest (.yaml/db3 files). Relative according to specified folder uri or path.

  • folder_uri – The uri location of the folder containing the rosbag db3 file and metadata yaml file. If not None, folder_path should be None.

  • folder_path – The local folder path of the folder containing the rosbag db3 file and metadata yaml file. If not None, folder_uri should be None.

  • session_uid – The UID of the session.

  • set_id – The set id to associate with the ingested rosbag.

Returns:

The ID of the added Rosbag2 asset.

Return type:

str

annotations_to_objects(ids=None, search=None, create_objects=False, margin_ratio=0.1, output_set_id=None, margin_pixels=None, source_annotation_type=None, session_uid=None, add_to_session=True, wait_for_completion=True, progress_callback=None)

Extracts objects from annotations.

Converts annotations to objects from annotations that have been previously ingested into SceneBox. Gets labels from the asset’s auxiliary metadata.

Parameters:
  • ids – IDs of the previously ingested annotations.

  • search – Query to locate the data subset of interest. Filters through existing assets according to the dictionary passed.

  • create_objects – If False and if mask annotations are available for the given asset id, adds the area of each annotation to the asset’s auxiliary metadata. Else, additionally creates object entities from the existing annotation data associated with the passed ID.

  • margin_ratio – Widens/shrinks the object’s area of interest. A larger number increases the object’s margin. Minimum value 0.

  • margin_pixels – Widens/shrinks the object’s area of interest by this amount. Minimum value 0. If both margin_rotio and margin_pixels are specified, margin_pixes takes precedence.

  • output_set_id – The name of the set to add the created objects to, if create_objects is True. Otherwise, has no effect.

  • source_annotation_type – If given, extracts objects from the annotation type specified. Otherwise, extracts objects from the any/all of the existing annotation types: polygons, bounding boxes, poses

  • session_uid – The session associated with the objects. Adds an event interval (and thus, an entity) to the existing session.

  • add_to_session – Whether to add the object to the session with session_uid for session search

  • wait_for_completion – If True, polls until job is complete. Raises an error if job fails to finish after (2 * #ids) seconds if IDs are supplied. Otherwise, continues execution and does not raise an error if the job does not complete.

  • progress_callback – Callback function to log the progress of the inference job. The callback should accept a parameter of type float for progress.

Returns:

ID of the Job that runs/ran the object extraction job.

Return type:

str

Raises:

ResponseError – If endpoint call does not get a valid response

get_annotation_sources(search=None)

Returns all sources of annotation for all images or images that satisfy a search query

Parameters:

search – Query to locate the data subset of interest. Filters through existing assets (images) according to the dictionary passed.

Returns:

a list of dictionaries that include provider, version, annotation_group, and annotation_type information for existing annotation sources

Return type:

List

get_annotation_labels(search=None)

Returns all labels for all images or images that satisfy a search query

Parameters:

search – Query to locate the data subset of interest. Filters through existing assets (images) according to the dictionary passed.

Returns:

a list of strings which are labels of existing annotation entities

Return type:

List

model_inference(asset_type='images', ids=None, search=None, model='mask_rcnn', obtain_mask=False, obtain_bbox=False, obtain_object_entities=False, obtain_embeddings=False, threshold=None, classes_of_interest=None, wait_for_completion=False, progress_callback=None, additional_params=None)

Perform inference from a list of supported models. See for supported models.

Extracts inferences such as masks, bounding boxes, objects, and/or embeddings on an asset of choice using a model of choice. Select from Mask RCNN, StarDist, or Image Intensity Histogram.

Parameters:
  • asset_type – The type of assets to perform model inference on. To ensure using a valid asset_type, you can import AssetsConstants from scenebox.constants

  • ids – IDs of the previously uploaded assets on which to perform a model inference.

  • search – Query to locate the data subset of interest. Filters through existing assets according to the dictionary passed.

  • model – The model to perform the inference with. Select from Mask RCNN (‘mrcnn’), StarDist (‘stardist’), or Image Intensity Histogram (‘histogram’)

  • obtain_mask – If True and model is Mask RCNN or StarDist, infer segmentations and add masks as a scenebox.models.annotation.SegmentationAnnotation. Otherwise, do nothing but log an error if the model chosen is Image Intensity Histogram.

  • obtain_bbox – If True and model is Mask RCNN or StarDist, infer bounding boxes from masks add the masks as a scenebox.models.annotation.BoundingBoxAnnotation. Otherwise, do nothing but log an error if the model chosen is Image Intensity Histogram.

  • obtain_object_entities – If True and model is Mask RCNN, obtains and adds object entities. Otherwise, do nothing but log an error if the model chosen is Image Intensity Histogram or StarDist.

  • obtain_embeddings – If True, extract asset embeddings with the chosen model.

  • image_size – If not None, resizes the input images to the provided size before performing inference(s). Otherwise, do nothing.

  • wait_for_completion – If True, polls until job is complete. Otherwise, continues execution and does not raise an error if the job does not complete.

  • progress_callback – Callback function to log the progress of the inference job. The callback should accept a parameter of type float for progress.

  • additional_params – A JSON formatted key-value pair structure for additional parameters such as threshold and classes of interest supported by the chosen model.

image_properties_enrichment(asset_type='images', properties=None, ids=None, search=None, wait_for_completion=False, progress_callback=None, additional_params=None)

Enrich images or objects with classical image processing features.

Enrich image or object assets with classical image properties/features such as brightness or contrast and add this as auxiliary metadata for the asset.

Parameters:
  • asset_type – The type of assets to perform enrichment on. To ensure using a valid asset_type, you can import AssetsConstants from scenebox.constants

  • ids – IDs of the previously uploaded assets on which to perform a model inference.

  • search – Query to locate the data subset of interest. Filters through existing assets according to the dictionary passed.

  • properties – A list of strings for each property to be added to the metadata of the asset. Choose from [‘mean_brightness’, ‘rms_contrast’, ‘variance_of_laplacian’].

  • wait_for_completion – If True, polls until job is complete. Otherwise, continues execution and does not raise an error if the job does not complete.

  • progress_callback – Callback function to log the progress of the inference job. The callback should accept a parameter of type float for progress.

  • additional_params – A JSON formatted key-value pair structure for additional parameters such as batch_size for the task.

extract_images_from_videos(video_id, extract_start_time=None, extract_end_time=None, fps=None, set_id=None, wait_for_completion=False)

Extract images from video

Parameters:
  • video_id – id of the video to extract images from

  • extract_start_time – optional start timestamp for extraction interval. If not given, will be set to start timestamp of the video.

  • extract_end_time – optional end timestamp for extraction interval. If not given, will be set to end timestamp of the video.

  • fps – desired fps (frame per second) value for extraction If not given, default value of 1.0 / 3.0 will be used.

  • set_id – The ID of the set to add the images to

  • wait_for_completion – If True, polls until job is complete. Raises an error if job fails to finish after (2 * #ids) seconds if IDs are supplied. Otherwise, continues execution and does not raise an error if the job does not complete.

Returns:

The ID of the Job that attempts to perform image extraction.

Return type:

str

extract_subclip_from_videos(video_id, subclip_id=None, extract_start_time=None, extract_end_time=None, set_id=None, wait_for_completion=False)

Create a subclip of a video.

Parameters:
  • video_id – The ID of the full-length video to extract from

  • subclip_id – The ID of the generated subclip

  • extract_start_time – The time in the full-length video to start the subclip from

  • extract_end_time – The time in the full-length video to end the subclip at

  • set_id – The ID of the set to place the generated subclip into

  • wait_for_completion – If True, polls until job is complete. Otherwise, continues execution and does not raise an error if the job does not complete.

Returns:

The job ID of the Job that tracks the subclip extraction.

Return type:

str

make_video_from_image_uris(bucket, folder, cloud_storage, sensor, session_uid=None, video_id=None, tags=None, set_id=None, image_skip_factor=1, wait_for_completion=False)

Make video from frames in a given cloud_storage, bucket, and folder. The path “{cloud_storage}://{bucket}/{folder}/” should contain frames with the following format: Unix_timestamps_in_nanoseconds.extension e.g. 1664999890113942016.jpeg

Parameters:
  • bucket – bucket where the frames are located

  • folder – folder where the frames are located. Note that bucket should not be included here.

  • cloud_storage – cloud provider e.g. s3 or gs

  • sensor – Sensor name associated with the image uris

  • session_uid – The ID of the session that the images and video belong to

  • video_id – The ID of the video

  • tags – Tags to add to video metadata

  • set_id – The ID of the set to add the video to

  • image_skip_factor – An integer number to use, if user wants to skip images when making video. Default is 1; skiping no image.

  • wait_for_completion – If True, polls until job is complete. Raises an error if job fails to finish after (2 * #ids) seconds if IDs are supplied. Otherwise, continues execution and does not raise an error if the job does not complete.

Returns:

The ID of the Job that attempts to perform video making.

Return type:

str

make_video_with_annotations_from_uris(timestamps_image_uris_csv_filepath, sensor, width, height, image_format, session_uid=None, video_id=None, tags=None, set_id=None, image_skip_factor=1, wait_for_completion=False)

Make video with overlayed annotations from provided image uris and annotaion uris per provider

Parameters:
  • timestamps_image_uris_csv_filepath – filepath for the csv file that contains two columns; timestamp and image_uri

  • sensor – Sensor name associated with the image uris

  • width – Width of images

  • height – Height of images

  • image_format – Format of images e.g. png

  • session_uid – The ID of the session that the images and video belong to

  • video_id – The ID of the video

  • tags – Tags to add to video metadata

  • set_id – The ID of the set to add the video to

  • image_skip_factor – An integer number to use, if user wants to skip images when making video. Default is 1; skiping no image.

  • wait_for_completion – If True, polls until job is complete. Raises an error if job fails to finish after (2 * #ids) seconds if IDs are supplied. Otherwise, continues execution and does not raise an error if the job does not complete.

Returns:

The ID of the Job that attempts to perform video making.

Return type:

str

similarity_search_bulk_index(embedding_ids=None, search_dic=None, wait_for_completion=False)

Bulk indexes embeddings to allow for similarity search.

Existing embeddings are mapped to a high-dimensional vector space. Then a distance measure is applied to find which assets are the most similar. After running this method, similarity search is available for the assets associated with the inputted embeddings.

Parameters:
  • embedding_ids – The IDs of the embeddings to bulk index.

  • search_dic – Query to locate the data subset of interest. Filters through existing assets according to the dictionary passed.

  • wait_for_completion – If True, polls until job is complete. Raises an error if job fails to finish after (2 * #ids) seconds if IDs are supplied. Otherwise, continues execution and does not raise an error if the job does not complete.

Returns:

The ID of the Job that attempts to perform bulk indexing.

Return type:

str

create_umap(asset_type, index_name, search_dic=None, n_neighbors=20, min_dist=0.3, train_size=10000, progress_callback=None, transform_only=False, wait_for_completion=False)

Applies UMAP onto existing asset indices.

Applies UMAP (Uniform Manifold Approximation and Projection) to existing asset embeddings, and outputs a visual representation of the embeddings space. Can be viewed from the “Embeddings” view on the SceneBox web app.

Note

Embeddings must be indexed before this function can be used.

Parameters:
  • asset_type – The type of asset to apply UMAP on. To ensure using a valid asset_type, you can import AssetsConstants from scenebox.constants

  • index_name – index_name of the embeddings. Used to filter through the embeddings metadata.

  • search_dic – Query to locate the data subset of interest. Filters through existing assets according to the dictionary passed.

  • n_neighbors – Controls how UMAP balances local vs. global structure in data. Low values of n_neighbours forces UMAP to concentrate on local structure, while high values push UMAP to look at larger neighbourhoods of each point.

  • min_dist – Controls how tightly UMAP is allowed to pack points together.

  • train_size – How much of the data should be used to train the umap. Default is 10000

  • progress_callback – Callback function to log the progress of the inference job. The callback should accept a parameter of type float for progress.

  • transform_only – If umap already exists and new data is needed to be transformed to the same embedding space, setting this parameter to True will take care of that.

  • wait_for_completion – If True, polls until job is complete. Otherwise, continues execution and does not raise an error if the job does not complete.

Returns:

The job_id of the Job that applies UMAP.

Return type:

str

search_similar_assets(id, similar_asset_count=10, asset_type='images', embedding_space=None)

Find top k similar assets to an existing asset with embeddings.

Parameters:
  • id – asset id.

  • similar_asset_count – count of similar assets (10 by default)

  • asset_type – type of assets (Images by default) To ensure using a valid asset_type, you can import AssetsConstants from scenebox.constants

  • embedding_space – embedding space in which similarity is performed on. Ineffective/optional if there is only one embedding space

Returns:

list of asset ids similar to the asset sorted by most to least similar.

Return type:

list

download_annotations(ids, folder_path=None)

Download annotations to a destination folder including the mask files

Parameters:
  • ids – list of annotation ids

  • folder_path – optional destination folder_path. If not given, all objects will be downloaded and returned as a result

Return type:

Dict

add_user(email, username, role, organization_id)

Add a new SceneBox user account to your organization.

Parameters:
  • email – E-mail of the user to add.

  • username – Username of the user to add.

  • role – Role of the user to add. Choose from admin, tester, health_checker, data_provider, data_user, or public_user. See the SceneBox user guides for more information.

  • organization_id – Organization id to add the user to.

Returns:

The successfully created user username, email, role, and organization.

Return type:

dict

modify_user(username, email=None, role=None, organization_id=None)

Modify an existing user account.

Please note that an existing user’s username cannot be changed.

Parameters:
  • username – The user username to modify.

  • email – The new user email.

  • role – The new user role. Choose from admin, tester, health_checker, data_provider, data_user, or public_user. See the SceneBox user guides for more information.

  • organization_id – The new user organization.

list_users()

Returns a list of all users.

Returns:

Returns a dictionary of username, organization name, role, email, and token for each user.

Return type:

List[Dict]

add_organization(organization_name)

Add a new SceneBox organization. Requires superadmin privilege.

Parameters:

organization_name – organization name.

Returns:

The successfully created organization id, name.

Return type:

dict

list_organizations()

Returns a list of all organizations.

Returns:

Returns a list of dictionary of organization name and id.

Return type:

List[Dict]

modify_organization(organization_id, organization_name)

Modify an existing organization. :Parameters: * organization_id – Organization id

  • organization_name – The new organization name.

list_ownerships()

Returns a list of all ownerships.

Returns:

A list of what ownerships the user associated with

Return type:

List[Dict]

display_image(image_id)

Display a given image on the SceneBox web app.

Note

This method will only run properly from a Google Colab or Jupyter notebook. Please also make sure that you are logged into the SceneBox web app, and that pop-ups are allowed in your browser.

Parameters:

image_id – The ID of the image to display on the SceneBox web app.

display_video(video_id)

Display a given image on the SceneBox web app.

Note

This method will only run properly from a Google Colab or Jupyter notebook. Please also make sure that you are logged into the SceneBox web app, and that pop-ups are allowed in your browser.

Parameters:

video_id – The ID of the video to display on the SceneBox web app.

display_object(object_id)

Display a given object on the SceneBox web app.

Note

This method will only run properly from a Google Colab or Jupyter notebook. Please also make sure that you are logged into the SceneBox web app, and that pop-ups are allowed in your browser.

Parameters:

object_id – The ID of the object to display on the SceneBox web app.

display_lidar(lidar_id)

Display a given Lidar on the SceneBox web app.

Note

This method will only run properly from a Google Colab or Jupyter notebook. Please also make sure that you are logged into the SceneBox web app, and that pop-ups are allowed in your browser.

Parameters:

lidar_id – The ID of the Lidar to display on the SceneBox web app.

display_session(session_uid)

Display a given session on the SceneBox web app.

Note

This method will only run properly from a Google Colab or Jupyter notebook. Please also make sure that you are logged into the SceneBox web app, and that pop-ups are allowed in your browser.

Parameters:

session_uid – The ID of the session to display on the SceneBox web app.

display_set(set_id)

Display a given set on the SceneBox web app.

Note

This method will only run properly from a Google Colab or Jupyter notebook. Please also make sure that you are logged into the SceneBox web app, and that pop-ups are allowed in your browser.

Parameters:

set_id – The ID of the set to display on the SceneBox web app.

display_projects(project_id)

Display a given project on the SceneBox web app.

Note

This method will only run properly from a Google Colab or Jupyter notebook. Please also make sure that you are logged into the SceneBox web app, and that pop-ups are allowed in your browser.

Parameters:

project_id – The ID of the project to display on the SceneBox web app.

display_dashboard(dashboard_name)

Display a given dashboard on the SceneBox web app.

Note

This method will only run properly from a Google Colab or Jupyter notebook. Please also make sure that you are logged into the SceneBox web app, and that pop-ups are allowed in your browser.

Parameters:

dashboard_name – The Name of the dashboard to display on the SceneBox web app.

clear_cache(all_organizations=False, partitions=None)

Clears scene engine’s Redis cache by organization Args:

all_organizations: if True, Redis cache for all organization is cleared. Requires Admin privilege partitions: list of partitions (like images, sets, annotations) to clear their cache, if not set, all is cleared

Returns:

ACK_OK_RESPONSE on success

jira_is_connected()

Check Jira connection :returns: is jira connected :rtype: bool

jira_create_issue(summary, type='Task', description=None, user_id=None, link_url=None, link_title='attachment')

create a Jira issue

Note

This method only works if SceneBox and Jira are connected

Parameters:
  • summary – ticket summary or title

  • type – ticket type Task | Story | Bug

  • description – ticket description

  • user_id – optional assignee

  • link_url – optional link url

  • link_title – optional title for link

Returns:

url of the created Jira ticket

Return type:

str

create a new Jira issue and link to an existing one

Note

This method only works if SceneBox and Jira are connected

Parameters:
  • existing_issue_key – Key for an existing Jira issue

  • summary – ticket summary or title

  • type – ticket type Task | Story | Bug

  • description – ticket description

  • user_id – optional assignee

  • link_url – optional link url

  • link_title – optional title for link

  • link_comment – optional comment for link

Returns:

url of the created Jira ticket

Return type:

str

save_campaign_operation_config(name, config, operation_id, description=None, id=None)

Create a config for a campaign operation.

Parameters:
  • name – Name of the configuration.

  • config – Body of the config to create. Form dependent on the configuration needs.

  • description – A brief description of the config’s purpose, settings, etc.

  • id – optional unique identifier

  • operation_id – String. The id of the campaign operation the config if meant for.

Returns:

The ID of the created config.

Return type:

str

delete_campaign_operation_config(config_id)

Delete a config.

Parameters:

config_id – The ID of the config to delete.

Raises:

ResponseError: – If the config could not be deleted.

get_campaign_operation_config(config_id, complete_info=False)

Get config metadata.

Parameters:
  • config_id – The ID of the config to receive the metadata of.

  • complete_info – If True, returns the the entire metadata of the config (with the config parameters contained inside). Otherwise, only returns the config parameters.

Returns:

The metadata and/or parameters of the desired config.

Return type:

dict

Raises:

ResponseError: – If the config metadata could not be obtained.

create_campaign(name, operations_pipeline, tags=None)

Create a Campaign A Campaign is a collection of tasks to run on the raw data or metadata on SceneBox. Use this method to create a new campaign with a template pipeline for the campaign. A template pipeline is the blueprint of an operations pipeline that every new task added to a campaign will follow. A pipeline is made up of TemplateOperation objects that define the order of the operations in a task along with their configs.

Parameters:
  • name – A name to associate the campaign with. Needs to be unique for an organization.

  • tags – List of strings. Attach a list of tags to a campaign to search / sort for them later.

  • operations_pipeline – List of TemplateOperation objects. A list defining the order of the operations with their respective configs.

Returns:

A unique id of the campaign if creation was successful.

Return type:

str

Raises:

CampaignError – If the campaign could not be created

delete_campaign(campaign_id)

Delete a campaign Deleting a campaign will delete all tasks and by extension all operation runs contained in it.

Parameters:

campaign_id – ID of the campaign to delete

Returns:

If the campaign could not be found or could not be deleted.

Return type:

ResponseError

search_campaigns(filters=None, return_fields=None, sort_field=None, sort_order=None)

Search for campaigns that satisfy the filtering criteria

Parameters:
  • filters – A list of key-value dictionaries to impose a criteria on available campaigns.

  • return_fields – List of strings. The fields to return from the available campaign metadata fields.

  • sort_field – String. The field to sort the return record with. Must be a valid field in the campaigns metadata. By default, results are sorted by their timestamps.

  • sort_order – String. One of [“asc”, “desc”] to indicate the order in which to return the results. By default, results are sorted in the descending order.

Returns:

A list of campaign details as a dictionary. A ResponseError if the task could not be found.

Return type:

List[dict]

add_task_to_campaign(campaign_id, input_payload=None)

Add new tasks to a campaign. Each new task will follow the blueprint pipeline of the operations and their configs defined as the campaign’s template task.

Parameters:
  • campaign_id – String. ID of the campaign to define the template for.

  • input_payload – Dictionary. Each dictionary key should conform to the input schema outlined for the first operation in the pipeline.

Returns:

ID of the task created

Return type:

str

Raises:

CampaignError – If the task could not be created successfully.

search_tasks(filters=None, return_fields=None, sort_field=None, sort_order=None, include_template_tasks=False)

Search for tasks that satify the filtering criteria

Parameters:
  • filters – A list of key-value dictionaries to impose a criteria on available tasks.

  • return_fields – List of strings. The fields to return from available task metadata fields.

  • sort_field – String. The field to sort the return record with. Must be a valid field in the tasks metadata. By default, results are sorted by their timestamps.

  • sort_order – String. One of [“asc”, “desc”] to indicate the order in which to return the results. By default, results are sorted in the descending order.

  • include_template_tasks – Boolean. Flag to indicate whether to search over template tasks set for a campaign.

Returns:

A list of task details as a dictionary. A ResponseError if the task could not be found.

Return type:

List[dict]

delete_task(task_id)

Delete a task Deleting a task will delete operation instances associated with it.

Parameters:

task_id – ID of the task to delete

Returns:

If the task could not be found or could not be deleted.

Return type:

ResponseError

list_operations(return_fields=None)

Get specifications of an operation

This method can be used to get about the definition and specification of an operation listed in the OpsStore.

Parameters:

return_fields – List of strings. The fields to return from the operation spec.

Raises:

ResponseError – If the operation could not be found.

Return type:

dict

get_operation(operation_id)

Get specifications of an operation

This method can be used to get about the definition and specification of an operation listed in the OpsStore.

Parameters:

operation_id – String. A unique string identifier for the operation.

Raises:

ResponseError – If the operation could not be found.

Return type:

dict

run_operation_instance(operation_instance_id, phase=None, phase_payload=None, wait_for_completion=False)

Run an individual instance of an operation contained in a task. An operation instance is a self-contained work unit of a task. Running a task will consume the input defined in its input payload according to its pre-defined config.

Parameters:
  • operation_instance_id – String. ID of the campaign to define the template for.

  • phase – Optional string. An operation can have different run-time behaviours based on its phase. Refer to the operations Phase and Phase Payload Schema for details.

  • phase_payload – Optional dictionary. A parameter key-value dictionary to provide phase-level configuration to the operation. Will be validated against the phase_payload_schema defined in the operation.

  • wait_for_completion – If True, polls until job is complete. Otherwise, continues execution and returns job_id

Returns:

ID of the job running the operation.

Return type:

str

Raises:

OperationError – If the operation could not be run successfully.