Review and export¶

Review connects processing success to scientific usability.

Read the overview first¶

Confirm:

• job state
• input image count
• processed and remaining counts
• result or mask count
• analysis scope and point or mask method
• linked project or site visit

Review a defensible sample¶

Prioritize:

1. low-confidence or flagged outputs
2. rare classes
3. images from unusual depth, lighting, or habitat
4. disagreement between model and analyst expectation
5. a random sample of apparently straightforward outputs

Record who reviewed the output, the rule used for corrections, and any known limitations.

Export¶

Use the export action on the relevant job or source workspace. Before downstream analysis:

• preserve the original export unchanged
• record the export date and job name
• verify row, point, image, or mask counts
• retain identifiers needed to join results to images and visits
• document label mappings and excluded data