Quick start¶
This path takes you from sign-in to a reviewable analysis without requiring a custom model.
Before you begin¶
Prepare:
- your ReefNetAI account
- a project name and a short scientific objective
- the survey location and visit date
- reef images in a consistent orientation and usable resolution
- either a labelset, a CoralNet source ID, or a plan to create labels in ReefNetAI
1. Create the fieldwork structure¶
Open Projects and select New Project. Add a survey and a site visit beneath it.
Use names that remain meaningful in an export, for example:
Farasan 2026 → Spring bleaching survey → Abu Latt reef · 2026-05-18
2. Add source data¶
Open the site visit, select Assets, and upload the images from that visit. Confirm the ready count before starting analysis.
Open CoralNet Imports, import the source, and attach the completed dataset to your site visit. This copies a stable working snapshot into the visit.
3. Choose the analysis¶
| Scientific task | Start with |
|---|---|
| Estimate cover at points | Classification |
| Delineate colonies or benthic regions | Segmentation |
| Reuse CoralNet points | A linked classification or point-seeded segmentation job |
| Train a context-specific classifier | The site visit’s Source workspace |
4. Review before export¶
Open the job’s Results or Review Queue. Check low-confidence or unusual outputs against the original image and your field context. Export only after the job state and processed counts match your expectation.
Do not treat a completed job as scientific validation
Completed means processing finished. It does not mean every label or mask is correct.
Completion checklist¶
- Project, survey, site, and visit date are correct
- Expected images are present and readable
- The labelset matches the study question
- Point or mask provenance is understood
- Uncertain outputs were reviewed
- The exported row or mask counts are plausible