Every piece of unlabeled data you import into Labelbox gets its own Data Row. In addition to the asset to be labeled, you can include additional information on each Data Row to help you filter and manage your data in Labelbox.
Each Data Row may contain the following information.
At the very minimum, your JSON import file should include at least one URL to a cloud-hosted file.
Metadata is a set of reserved fields with predefined names and types. Fields are
embedding. Visit Metadata to see a JSON sample.
Any Data Row can have metadata. In order to keep the metadata fields consistent across all of your Datasets, all Data Rows within an organization will reference a single metadata schema, allowing you to use a consistent set of fields to search across all of your datasets.
You can use metadata fields to explore your datasets across the Labelbox platform for the purpose of active learning and model error analysis. This powerful exploration tool can help you identify patterns in your datasets, assist with dataset selection for labeling, and help with quality control. Additionally, the Model Diagnostics workflow enables you to filter by metadata to develop a deeper understanding of how your training data can be improved.
Media attributes include the following fields:
mimeType. Visit Media attributes to see a JSON sample.
Media attributes are additional information about each asset that helps you filter and search across your datasets.
Attachments are useful for providing additional context for your labeling team in the Editor. Attachments include the following types:
IMAGE_OVERLAY. Visit Attachments to see a JSON sample.
You will be able to view
HTML attachments in the right panel in the Labelbox Editor. If your Data Row has attachments, you'll be able to click on the attachments icon in the top left corner of the Editor and see the attachments will pop out in the right panel.
Support for HTML attachments will allow you to provide richer resources to labelers in the attachments panel. The new HTML type can take a text or HTML file and render it in an iframe in the attachments panel. This allows your labeling team to view HTML attachments in the attachments panel instead of needing to click on a URL to view it. You can also use markdown/HTML to style the content.
Or if your Data Row has an
IMAGE_OVERLAY attachment, you'll be able to toggle through the additional image layers in the Editor canvas. Click on the layers icon in the top right corner of the Editor to view the image layers.
Updated 18 days ago