Metadata is non-annotation information about the asset to be labeled. You can use metadata to search and filter across your data rows in Labelbox.
There are two types of metadata: reserved fields (user cannot change) and custom (user-defined) fields.
The metadata schema lives at the organization level, allowing you to apply the same metadata fields across multiple datasets. To view your metadata schema, go to the Schema tab and click on the Metadata subtab.
Attachments vs metadata
In an earlier version of Labelbox, what is now called attachments used to be called metadata. Attachments provide additional context for labelers but are not searchable within Catalog.
Upon upload, Labelbox automatically defines a set of reserved fields on your data rows to help you easily access the full range of features in Labelbox. Each metadata field has a unique schema ID that is used to upload data to Labelbox.
|Free text field|
|An embedding field|
|Enum - [train, valid, test]|
|ISO 8601 datetime field. All times must be in UTC|
|Embedding computed for uploaded image data|
|Embedding computed for uploaded text data|
All metadata is strictly typed. These are the custom fields Labelbox offers.
|Used for similarity. A float vector of length 128||Similarity|
|free text field. Max 1024 characters.||Equals & prefix matching|
|Enum field with options.||Equals|
|Option of an enum. Max 64 options can be created per Enum type. 128 for enterprise customers and can be further increased upon request.||Equals|
|An ISO 8601 datetime field. All times must be in UTC timezone||Equals, greater than, less than, between|
|Integer or Float||Equals, greater than, less than, between|
Yes, labelers can see metadata from the data row information panel. For more details, please refer to the page Data row Information panel.
Not directly. But you could use the Catalog to query a set of data rows that have specific metadata in common and add to the project for labeling using Batch queue (beta) .
Updated 18 days ago