Simplify workflows with the Labelbox SDK

Just because Labelbox has a beautiful and intuitive UI doesn’t mean you need to use it for everything. Every aspect of Labelbox is engineered for speed, and that includes the most recent improvements we’ve made to our SDK. It’s now easier than ever to get started with our SDK thanks to simplified documentation and a set of new tutorials built in end-to-end runnable Google Colab notebooks and GitHub. With the help of these tutorials, you can jump right into programmatically managing projects, users and user roles, and running model-assisted labeling pipelines.

Learn more about the SDK and make sure you’re running the latest version for full functionality.

Get started with powerful tutorials

We’ve made new tutorials to help you get started with the SDK quicker than ever. With live notebooks in Github and Google Colab, you can run tutorials with your own data without having to first configure them to work with your projects. We know learning and troubleshooting new SDKs can be time consuming and distracting, but we’re confident that live tutorials will help you onboard in no time by seeing working examples of common operations.

Set up new projects faster than ever

Creating and configuring new projects can be done faster than ever through the updated SDK. Users can programmatically attach datasets and upload labeling instruction PDFs. You can also build your ontology with Python objects instead of using JSON which will save time and is less likely to result in ontology errors.

Customers are constantly creating simple yet powerful scripts to speed up their operations. For example, some customers want to label a single dataset multiple times for different purposes. They can easily make a script that accepts different tool names as parameters and every other part of project setup is completely automated, saving time and reducing the likelihood of configuration errors.

Invite users and assign project roles

We made managing a team in Labelbox more powerful in the UI earlier this year, but you can now use the SDK to programmatically manage users and roles. Invite teams to join your organization and set organization roles with a short and simple script. Assign users to new projects and manage their label or review queue with a short string of commands. This can be a significant time savings for organizations as they regularly spin up new projects for experimentation and iteration.

Streamline and accelerate model-assisted labeling

Running a model-assisted labeling pipeline just got easier. We’ve improved MAL error validation to identify common errors like invalid feature schema IDs, invalid data row IDs, and incorrectly formatted uploads to name a few. Error validation not only tells you what went wrong, but it surfaces the error before uploading the predictions so you can more quickly identify and correct errors.

We’re always looking for ways to make workflows easier and more powerful. Through the SDK, you can now pull a list of all queued data rows within a project so you can more quickly upload model inferences for large datasets.

Finally, MAL now works on tiled imagery datasets so you can put it to use with geospatial data.

There’s more

In addition to these updates, we’ve made a few other improvements you should know about:

  • We’ve added input validation for webhooks and queue prioritization. This will help make error messages more clear and leave less room for guesswork.
  • You can now bulk export issues and comments. This can help you identify and skip exporting data rows that have outstanding issues during training.
  • It’s now easier to check the status of bulk import requests. Status, errors, and inputs are now properties of the BulkImportRequest.

Get started today

All of these updates are ready to use today but you’ll need to make sure you’re using the latest version of the SDK for full functionality. You can always find the latest version on GitHub, along with all of our tutorials and other configuration support resources.

Tags

Labelbox

Labelbox is a collaborative training data platform empowering teams to rapidly build artificial intelligence applications.