How saikumar-2210/marine_sight works
MarineSight turns Sentinel imagery into oil-spill detections, verifies likely slicks with optical and metocean evidence, attributes likely vessels, and presents scans in an operations map. Operators choose a date or request a rescan; the API serves stored results or starts a scan, persists job/results, and streams progress to the UI. The graph preserves the documented AIS, coverage/tiles, scheduling, and evidence-map capabilities. Pipeline calls and UI/API wiring are cited where shown; unsampled server and model internals are represented at their documented subsystem paths.
Main entry points: pipeline.py Date scan pipeline · app.js HTTP and event API · OperationsDashboard.jsx Operations dashboard · scanJobs.js Scan job coordinator
Components 31
Detection pipeline
- Date scan pipeline
ml_service/marinesight/pipeline.py - Sentinel scene catalog
ml_service/marinesight/cdse.py - Geospatial raster grid
ml_service/marinesight/raster.py - Slick segmentation
ml_service/marinesight/segment.py - U-Net inference
ml_service/marinesight/unet.py - Optical verification inputs
ml_service/marinesight/optical.py - MetOcean features
ml_service/marinesight/metocean.py - Oil probability verifier
ml_service/marinesight/verifier.py - AIS drift attribution
ml_service/marinesight/attribution.py - SAR vessel detection
ml_service/marinesight/vessels_sar.py
Scan services
- HTTP and event API
server/app.js - Scan job coordinator
server/scanJobs.js - Daily scan scheduler
server/scheduler.js - SQLite scan and AIS data
server/db.js - AIS feed and history
server/ais.js - Sentinel coverage
server/sentinelCoverage.js - Sentinel map tiles
server/sentinelTiles.js
Operations interface
- Operations dashboard
src/pages/OperationsDashboard.jsx - Date and spill state
src/context/SpillContext.jsx - Marine API client
src/services/marineApi.js - Live scan events
src/hooks/useServerEvents.js - Map and evidence layers
src/components/map/MapContainer.jsx - Incident inspector
src/components/inspector/IncidentInspector.jsx - AIS vessel display
src/components/map/VesselLayer.jsx - Vessel taxonomy
src/utils/vesselTaxonomy.js
Model development
- Training data builder
ml_service/training/build_dataset.py - U-Net training
ml_service/training/train_unet.py - Verifier training
ml_service/training/train_verifier.py
Other components
- Operator
- Copernicus Data Space
- AIS stream
Connections 35
- Operator → Operations dashboard: uses
src/pages/OperationsDashboard.jsx - Operations dashboard → Date and spill state: reads state
src/pages/OperationsDashboard.jsx - Date and spill state → Marine API client: requests scans
src/context/SpillContext.jsx - Date and spill state → Live scan events: subscribes to
src/context/SpillContext.jsx - Operations dashboard → Map and evidence layers: renders
src/pages/OperationsDashboard.jsx - Operations dashboard → Incident inspector: renders
src/pages/OperationsDashboard.jsx - Map and evidence layers → Marine API client: fetches vessel tracks
src/components/map/MapContainer.jsx - AIS vessel display → Vessel taxonomy: classifies with
src/components/map/VesselLayer.jsx - Marine API client → HTTP and event API: HTTP requests
src/services/marineApi.js - Live scan events → HTTP and event API: WebSocket events
README.md - HTTP and event API → Scan job coordinator: routes scan requests
server/app.js - HTTP and event API → SQLite scan and AIS data: queries results
server/app.js - HTTP and event API → AIS feed and history: serves AIS status
server/app.js - HTTP and event API → Sentinel coverage: routes coverage
server/app.js - HTTP and event API → Sentinel map tiles: routes tiles
server/app.js - Daily scan scheduler → Scan job coordinator: starts daily scans
README.md - AIS feed and history → AIS stream: receives positions
README.md - AIS feed and history → SQLite scan and AIS data: stores history
README.md - Scan job coordinator → Date scan pipeline: spawns per date
README.md - Scan job coordinator → SQLite scan and AIS data: persists scan results
README.md - Date scan pipeline → Sentinel scene catalog: queries scenes
ml_service/marinesight/pipeline.py - Sentinel scene catalog → Copernicus Data Space: requests imagery
README.md - Date scan pipeline → Slick segmentation: segments scenes
ml_service/marinesight/pipeline.py - Slick segmentation → U-Net inference: runs inference
ml_service/marinesight/segment.py - Slick segmentation → Geospatial raster grid: uses grid operations
ml_service/marinesight/segment.py - Date scan pipeline → Optical verification inputs: fetches verification inputs
ml_service/marinesight/pipeline.py - Date scan pipeline → MetOcean features: fetches verification inputs
ml_service/marinesight/pipeline.py - Date scan pipeline → Oil probability verifier: scores candidates
ml_service/marinesight/pipeline.py - Date scan pipeline → AIS drift attribution: ranks likely culprit
ml_service/marinesight/pipeline.py - Date scan pipeline → SAR vessel detection: detects SAR vessels
ml_service/marinesight/pipeline.py - Training data builder → U-Net training: supplies training chips
README.md - Verifier training → Oil probability verifier: produces model
README.md - U-Net training → U-Net inference: produces weights
README.md - Sentinel coverage → Copernicus Data Space: searches coverage
README.md - Sentinel map tiles → Copernicus Data Space: loads imagery tiles
README.md