Import from CoralNet¶

CoralNet import brings a reusable source dataset into ReefNetAI. It can contain images, point annotations, labels, and metadata returned by CoralNet.

Import a source¶

1. In CoralNet, open the source and copy its numeric source ID from the URL.
2. In ReefNetAI, open CoralNet Imports.
3. Select Import CoralNet source.
4. Provide the source ID and the requested access when prompted.
5. Leave the page while the import runs; return to the dataset library to check status.

ReefNetAI requests available annotation scopes from CoralNet. You do not choose all or confirmed during import. A later workflow may let you choose a landed subset.

Confirm what landed¶

Open the dataset and check What landed and Artifact inventory.

Verify:

• image count
• point-annotation count
• label count
• returned scopes
• completed artifact rows
• any empty confirmed subset

Attach the dataset to fieldwork¶

Use one of the primary actions:

• Attach to SiteVisit — connect the latest successful snapshot to an existing visit.
• Create new SiteVisit from dataset — select a project, survey, and site, then create a visit around the imported data.

The attachment pins a particular import snapshot. A later CoralNet update does not rewrite that visit.

Check for updates¶

Select Check CoralNet for updates when you want to compare the local dataset with its source. If a new successful snapshot lands, older attachments may show stale.

For a stale attachment, choose Create updated SiteVisit to create a successor pinned to the new snapshot. The original visit remains unchanged for reproducibility.

Use imported points¶

From an attached site visit, create a linked job:

• Classification: choose Use imported CoralNet points.
• Segmentation: choose a point-seeded workflow, then Use imported CoralNet points.

CoralNet points are source observations; they are not segmentation-mask ground truth.