Developer guide: Export model run data via Python SDK

There are three ways you can export data from Labelbox: export from Catalog, export from a model run, and export from a labeling project. This guide explains how to export data from a model run.

If you choose to train a model in your custom ML environment outside of Labelbox, follow the instructions below to learn how to export your data from a model run. Alternatively, you can train a model in the Labelbox UI using the model training integration (see How to train a model in Labelbox).

How to export via the app UI

To export annotations from a model run via the UI, go to Model > select a model from the My models section > Model runs tab. You can use the filters to query data row label status, metadata, batch, annotations, and workflow history. Note that excluding fields from your export will make the export process faster and the export file smaller.

To export your annotations via the Labelbox UI, follow these steps:

Select a model run.
Configure your export

Option 1: Export all data rows

Click on All (X number of data rows), then click the Export data v2 (beta) in the drop-down menu.

Option 2: Export a split or a slice

On the left side of the Model Run page, click on one of the Split options (Train, Validation, or Test) or Slice options. Click on Slice/Split (X number of data rows), then click the Export data v2 (beta) in the drop-down menu.

Option 3: Filter data rows

You can also build your filters in the Model filters to query for data rows of the desired context.
Then click the Done (X number of data rows) dropdown, and click Export data v2 (beta) to export the data rows.

Option 4: Select specific data rows

Hand-select data rows to export using the check boxes next to each data row.
Then, click the drop-down on the right side X selected, click Export data v2 (beta) to export the data rows.

After you click Export JSON, you will see a notification banner telling you that you can track the progress of the export job in the Notifications center.
Once the job is complete, download the export file by clicking the Download link.