MineralVision

The premiere platform for mineral exploration, discovery, and compliance.

MineralVision unifies the full exploration lifecycle — satellite anomaly detection, target ranking, drilling, resource estimation, and compliant reporting — behind a single FastAPI backend and a single React web application. Read docs/VISION.md for the platform vision and the personas it serves.

Quick Start

API (canonical entry point)

pip install -r requirements.txt
cd MineralVision_Final_Package
uvicorn src.api.main:app --reload --host 0.0.0.0 --port 8000

Interactive API docs are then available at http://localhost:8000/docs (Swagger UI) and http://localhost:8000/redoc (ReDoc).

Note: src.api.main:app is the canonical application entry point. The historical main_demo.py / main_production.py / main_simple.py / main_standalone.py entry points are being consolidated into it.

Web UI

cd MineralVision_Final_Package/src/ui/web/mineralvision-app
npm ci
npm run dev

Docker

docker compose up --build

Features

• Drillhole database — collars, surveys, assays, lithology, with 3D visualization and cross-sections
• QAQC — standards, blanks, and duplicates tracked from assay load to resource report
• Geostatistics — variography, kriging, block modeling, grade shells (src/api/geostatistics/)
• Geophysical inversion — gravity, magnetics, EM (src/api/geophysics/)
• Prospectivity ML — target ranking with spatial cross-validation and uncertainty quantification (src/api/ml/)
• WALDO field detection — YOLO11 + RF-DETR ensemble detection of outcrop, gossan, and alteration indicators in field imagery
• Sensor fusion — magnetometry, radiometrics, hyperspectral, LiDAR, GPR, and SEG-Y with Kalman and deep-learning fusion (src/api/sensor_fusion/)
• Compliant reporting — JORC- and NI 43-101-aligned report generation with end-to-end audit trails (src/api/reporting/)
• Auth & multi-tenancy — JWT authentication with role-based access control

Project Structure

mineralvision/
├── MineralVision_Final_Package/     # Canonical API + UI
│   └── src/
│       ├── api/                     # FastAPI application (entry: src.api.main:app)
│       └── ui/web/mineralvision-app # React + TypeScript web app
├── MineralVision_Enhanced/          # Lakehouse + geospatial middleware
├── MineralVision_WALDO_Production_Package/  # WALDO detection services
├── infrastructure/                  # Kubernetes, Helm, Terraform
├── tests/                           # Pytest test suite
└── docs/                            # Documentation (start with VISION.md)

Oil-Spill Intelligence Extension

MineralVision now supports reviewable oil-spill assessment from drone, satellite, fluorosensor, or manual-annotation evidence. The extension converts a versioned segmentation mask into a cleaned oil footprint, area estimate, severity screen, GeoJSON geometry (when a geographic image footprint is supplied), persistent incident record, human review state, and advisory drone-coverage grid. Raw-image inference is disabled by default and requires an explicitly configured, trusted local TorchScript or ONNX model.

Read the research review, architecture and operating constraints, production roadmap, and custom model/API integration guide. The API surface is available under /api/oil-spill.

Development

# Backend tests
pytest tests/

# UI type-check / build (from the UI directory)
npm run build

License

Proprietary