Proof of Dirt is an open-source verification layer. Contributors can work on remote sensing, field systems, livestock AI, dashboards, and Solana oracle records — without turning the project into a token sale or custody product.
SAR plowing detection, NDVI yield forecasting, crop-stage classification, field-boundary validation, cloud/noise handling, and model benchmarking.
Offline-first agent app, plot GeoJSON, contract capture, RFID intake, livestock body-condition photos, inspections, and review workflows.
Solana record program, event schema, privacy-preserving hashes, indexers, dashboards, and integrations for institutional or RWA partners.
Sentinel-1 radar change detection for small and medium plots, with model accuracy reported by plot-size band and geography.
Sentinel-2 time-series models for pre-harvest yield forecasts, crop stress, anomaly detection, and harvest-readiness signals.
Body-condition scoring from field photos on mid-range Android phones, linked to RFID and veterinary records.
Financing milestones, verification attestations, settlement events, and evidence hashes on Solana. No tokens, no transfers — pure verification.
Offline-first field data capture for plot boundaries, contracts, RFID intake, inspections, photos, and dispute evidence.
Financier, operator, contributor, and public transparency views for verification events, confidence status, bounties, and reports.
Every funded bounty has a GitHub issue, written spec, acceptance criteria, reviewer, and reward amount.
Build to the spec, include tests or benchmarks where possible, and document the model, schema, or workflow change.
Funded bounties can be paid in USDC or SOL from the public treasury. Unfunded issues are labeled volunteer or research.
Open datasets and model challenges focus on SAR plowing detection, NDVI yield forecasts, and body-condition scoring. Winning models improve verified metrics and ship open source.
| Challenge | Target | Why it matters |
|---|---|---|
| SAR plowing detection | F1 by plot-size band | Small-plot verification is the frontier for smallholder finance. |
| NDVI yield prediction | RMSE under 15% where data supports it | Yield forecasts improve underwriting and early warning. |
| Livestock BCS | Low-error score from mobile photos | Animal condition evidence supports feed finance and fattening cycles. |
Volunteers and interns work on scoped deliverables with public contributor credit, technical review, and recommendation support where appropriate. Strong contributors get first consideration when funded roles open.