How ReefNetAI is organized¶
ReefNetAI separates field context from processing jobs so your results remain interpretable later.
Fieldwork hierarchy¶
Project → Survey → Site visit → Images & source data → Analysis jobs → Reviewed exports
| Object | Use it for | Example |
|---|---|---|
| Project | A research program or study | Red Sea bleaching response 2026 |
| Survey | A campaign or sampling design within a project | Post-summer belt transects |
| Site | A reusable geographic location | Abu Latt reef |
| Site visit | Observations at one site and date | Abu Latt · 2026-09-14 |
| Job | One classification or segmentation task | 50 random points per image |
Dataset, snapshot, and working copy¶
A CoralNet imported dataset is a reusable record of an external source. Each successful import creates an immutable snapshot. Attaching that snapshot to a site visit creates a working copy governed by the project.
This matters because:
- updates in CoralNet do not silently change an existing analysis
- the imported label snapshot preserves origin provenance
- site-visit label edits do not rewrite the import record
- one source can support more than one visit or analysis context
Linked and standalone jobs¶
Linked jobs belong to a site visit and inherit its field context and project sharing. Use them for durable research workflows.
Standalone jobs are useful for a quick, independent batch. They use the platform’s assistive model and do not own or train a custom source model.
Tip
If the work may contribute to a custom classifier, comparison across visits, or a site-level export, start from a site visit.