Working with Labels (API)
A label row has the attributes label_hash
uid, data_title
, data_type
, data_units
, and label_status
.
A label row contains a data unit or a collection of data units and associated labels, and is specific to a data asset with type video or image group.
-
A label row with a data asset of type
video
contains a single data unit. -
A label row with a data asset of type
img_group
contains any number of data units.
A data unit can have any number of vector labels (e.g. bounding box, polygon, keypoint) and classifications.
Getting label rows
Project label row IDs (label_hash
uid) are found in the project information, which also contain information about the data title (data_title
), data type (data_type
) and label status (label_status
).
Saving label rows
Labels are saved to their label row ID (label_hash
uid) from a label row instance. In case you want to save labels for the data which was not labeled before, follow the steps under “Creating label rows” below.
Label rows have to be saved in the same format as fetched. Click here for information on the label row structure.
Creating a label row
If you want to save labels to a unit of data (video
, img_group
) for which a label row (and thus a label_hash
uid) does not exist yet, you need to create a label row associated with the data.
-
Get the data_hash (
data_hash
uid) that you want to create labels for. For this, request all label rows and note the ones that are NOT_LABELLED under ‘label_status’ (or, wherelabel_hash
is None) following the instructions here. -
Create the label row:
The request returns a label row instance.
Was this page helpful?