Large Taxonomies in Label Studio 1.10.0 — November 2023 Community Newsletter

🌿 Large Scale Taxonomies in Label Studio 1.10.0!

The latest version of Label Studio is here, featuring the beta release of Large-Scale External Taxonomies! This new feature allows you to securely load, manage, and view taxonomies of hundreds of thousands of choices using a single JSON flat file.

Other improvements include an update to Label Studio dependency management that will make it easier to install Label Studio into your preferred MLOps environment and improvements to the Ranker interface that make it easier to fine-tune LLMs and QA bots with large numbers of generated answers.

Label Studio 1.10.0 also includes a number of security and bug fixes, and we encourage you to upgrade to take advantage of the new improvements and fixes. Head over to GitHub for the latest release notes, and check out Label Studio docs for how to install or update to the latest version.

🎢 Adala 0.0.4 Released!

The latest release of Adala is now available! Adala is a new open source framework for implementing agents specialized in data processing, with an emphasis on diverse data labeling tasks. The latest Adala release features significant improvements to the quality of automated prompt engineering.

If you’re new to agents and automated data labeling, check out the introductory blog post, "Our Vision For the Future of Reliable Labeling Agents.” You can learn more about this release and try out the new code generation and validation environment at the Adala GitHub page.

🧩 Segment Anything x Label Studio in Hugging Face Spaces!

Earlier this month, the Label Studio community released a new machine-learning backend to Hugging Face Spaces: Segment Anything! The new Space delivers the groundbreaking power of Segment Anything’s generalized object detection knowledge, using the MobileSAM build of the model as a Hugging Face Spaces application that will make it even easier to launch and connect to your Label Studio instance.

Segment Anything “is a promptable segmentation system with zero-shot generalization to unfamiliar objects and images, without the need for additional training.” This sort of generalization, combined with powerful keypoint and bounding-box cueing, makes SAM an ideal tool for pre-annotating image segmentation datasets. Read more about it in the Label Studio blog, and let us know what sort of machine-learning backend you’d like to see!

🎉 Community Shoutouts

• Special thanks to Alex Brown, representing Elttam, for their responsible security disclosures, helping to make Label Studio more secure for the entire community.
• Thanks to @davidblom603 for opening issue #4844 and @WillieMaddox for opening issue #5032, both of which helped us identify and fix a bug related to upgrading Label Studio.

🏗️ Help Grow the Label Studio Community

Do you have an integration, user story, or personal experience you want to share with the rest of the community? Reach out to us in the community Slack channel. We're especially interested in sharing your experiences with Machine Learning integrations.

Happy Labeling!