October 4, 2022

Release notes

App

  • We now have much better charts in Model. It helps users understand where their models are performing well and poorly, and it helps users prioritize what to label next. In the Model metrics view, there are model metrics recapped at the top of the Model Run, a binary confusion matrix, and a precision-recall curve. In the detailed view of Models, we now display model metrics to help you understand how well a model is performing on the specific Data Row.

  • Global keys are the new user-specified unique ID for your Data Rows. It has several advantages over external ids (external IDs are still supported). Global keys are unique on a Catalog / Organization level and each global key will only map to exactly one Data Row. This makes querying Data Rows much faster.

  • Workflows (beta) is enabled for Free/EDU tiers. You can use Workflows to create rule-based review tasks and multi-step sequences to reduce cost and increase the quality and efficiency of your labeling operations.

  • Our Document editor is now a multimodal annotation platform. Rendering the power of Bounding Box + NER, we now support a multitude of features in the document editor: raw data in export, TokenIDs/ability to select a token, and text layers rendered on top of the PDF.

  • The gallery view in Model now has infinite scroll, meaning you no longer need to click through pages.

  • Catalog now has an Attachment panel that enables you to visualize Attachments directly in Catalog. It works with special characters, images, videos, etc.

  • In Model, customers can now see Entities/NER (name entity recognition) annotations and predictions on Text data rows.

  • In Catalog, you can now save and reuse filters by saving your filters as a “Slice”. This means you no longer have to create the same filters over and over again. These slices are dynamic, meaning any incoming Data Row that matches the Slice filters will show up in the Slice.

  • In Catalog, you can now see the distribution of Annotations & Metadata in your data. This will help you understand the distribution of your data in Catalog and decide what to label next.

  • You can now view a Data Row's information in the Editor (e.g., external ID, asset type, created on, dataset, and attachments). These vary across data types and allow you to view specific media attributes, such as an image’s width and height, the number of pages in a PDF document, and the frame rate and duration of a video.

  • MAL and Labelimport are now supported on PDF bounding boxes for document users.

Python SDK

The latest version of our Python SDK is v3.27.2. See our full changelog in Github for more details on what was added recently.

Version 3.27.2 (2022-10-04)

Added

  • Added deprecation warning for missing media_type in create_project in Client.

Changed

  • Updated docs for deprecated methods _update_queue_mode and get_queue_mode in Project
  • Use the queue_mode attribute in Project to get and set the queue mode instead
  • For more information, visit https://docs.labelbox.com/docs/migrating-to-workflows
  • Updated project.export_labels to support filtering by start/end time formats "YYYY-MM-DD" and "YYYY-MM-DD hh:mm:ss"

Version 3.27.1 (2022-09-16)

Changed

  • Removed client.get_data_rows_for_global_keys until further notice

Version 3.27.0 (2022-09-12)

Added

  • Global Keys for data rows
    • Assign global keys to a data row with client.assign_global_keys_to_data_rows
    • Get data rows using global keys with client.get_data_row_ids_for_global_keys and client.get_data_rows_for_global_keys
  • Project Creation
    • Introduces Project.queue_mode as an optional parameter when creating projects
  • MEAToMALPredictionImport class
    • This allows users to use predictions stored in Models for MAL
  • Task.wait_till_done now shows a warning if task has failed

Changed

  • Increase scalar metric value limit to 100m
  • Added deprecation warnings when updating project queue_mode

Fixed

  • Fix bug in feature_confusion_matrix and confusion_matrix causing FPs and FNs to be capped at 1 when there were no matching annotations

Version 3.26.2 (2022-09-06)

Added

  • Support for document (pdf) de/serialization from exports
    • Use the LBV1Converter.serialize() and LBV1Converter.deserialize() methods
  • Support for document (pdf) de/serialization for imports
    • Use the NDJsonConverter.serialize() and NDJsonConverter.deserialize() methods