- When creating a video ontology via the UI or SDK, you can now add free-form text classifications at the global level.
- We released a new version of our export format, Export v2. Export v2 gives you more granular control over your data exports, allows you to include more data fields in your exports, and has overall better performance (read this guide for more details). You can now export from Labelbox in three ways:
- Labelers will be shown their next task in the labeling queue within 100ms.
- The width on the left toolbar in the editor is now adjustable.
- You can now use Catalog to classify a large amount of data in just a few clicks. Do this by curating a subset of data and then classifying them all simultaneously. To learn more, read bulk classification.
- You can now use common undo and redo keyboard shortcuts for raster masks in the DICOM editor. This is supported for masks created within an active session. To learn more, see our shortcuts page.
- The new Updates section in the app brings all communications related to a project into a single place. The Updates tab is designed to reduce the need for manually gathering information from disparate sources. To learn more, see Updates.
- Free-tier users are now restricted to 10,000 LBUs.
- Labelers can now view, update, and delete other users' labels in addition to their own. This makes the workflows feature on par with the legacy delete and requeue workaround. To learn more, see Rework in workflows.
- We will sunset the export v1 format by the end of 2023.
The latest version of our Python SDK is v3.46.0. See our full changelog in Github for more details on what was added recently.
- Global key support to DataRow Metadata
- Removed dataset-based projects from project setup notebook
- Updated all links to annotation import and prediction notebooks in examples README
- Reduce threshold for async batch creation to 1000 data rows
- Added subclassifications to ontology notebook
- Added conversational and pdf predictions notebooks
predictionsparam for optionally exporting predictions in model run export v2
- Limits on
project_idson catalog export v2 params
data_row_idsfilter for dataset and project export v2
- ISO timestamp parsing for datetime metadata
- Docstring typo for
- Removed mention of embeddings metadata fields
- Fixed broken colab link on
- Added free text classification example to video annotation import notebook
- Updated prediction_upload notebooks with Annotation Type examples
- Nested object classifications to
- Relationship Annotation Types
model_run_idsto params in all export_v2 functions
- VideoMaskAnnotation annotation import
- Added DICOM annotation import notebook
- Added audio annotation import notebook
- Added HTML annotation import notebook
- Added relationship examples to annotation import notebooks
- Added global video classification example
- Added nested classification examples
- Added video mask example
- Added global key and LPOs to queue management notebook