Crop Damage Data Collection Hackathon

FLock.io × UNDP Dominican Republic Crop Damage Data Collection Hackathon Powered by AI Arena on train.flock.io Tagline Capture the reality of your fields. Help build sovereign, decentralized AI that protects Dominican smallholder farmers from crop damage. About the Hackathon FLock.io, in partnership with UNDP Dominican Republic, is launching a local data-gathering hackathon with Dominican Republic universities and student communities. This hackathon directly delivers the data layer of FLock’s first end-to-end production deployment of decentralized AI for parametric crop insurance (as outlined in the FLock × UNDP DR Internal Alignment document). Students become active co-creators, not passive data sources. You will photograph and label real-world crop images from the exact communities that will benefit from the AI-powered insurance app. High-quality contributions will feed straight into train.flock.io / AI Arena — FLock’s decentralized Kaggle-style platform — where the global ML community will compete to train the crop-damage assessment model. The winning model will be deployed on FLock’s API platform, power the UNDP insurance app, and enable smart-contract-triggered payouts on Base. This is decentralized AI done right: data crowdsourced from the ultimate beneficiaries → trained openly with guaranteed rewards → serving real people with invisible crypto rails underneath. Why participate? - Earn cash prizes for the highest-quality contributions - Contribute directly to a live government-backed insurance project that helps farmers recover faster after storms - Be featured in FLock × UNDP case studies, press, and investor materials (Proof of Pitch Paris, Rwanda pipeline, etc.) - Gain hands-on experience with real-world decentralized AI and federated-learning workflows - Potential follow-on opportunities (internships, continued data roles, or participation in AI Arena training bounties) The Challenge Go into the field (farms, cooperatives, university plots, partner networks of AGRODOSA / CONACADO / ADOPEM, etc.) and capture high-quality smartphone images of crops under diverse real-world conditions. Required Classes (binary labeling) - Label 0 – Non-damaged (healthy/normal crops) - Label 1 – Damaged crops (visible storm, wind, flood, hail, pest, disease, or other weather-related damage typical for parametric insurance triggers) Key Requirements for High-Quality Data - Images must show clear focus on the crop/plot (leaves, stems, whole plants, or field-level view) - Capture diversity across: - Locations (different provinces, soil types, farms, cooperatives) - Backgrounds and conditions (sunny, cloudy, rainy, dawn/dusk, different growth stages, angles, distances) - Crop types common in DR (bananas, cacao, rice, coffee, etc.) - Minimum resolution: 1080p recommended (higher is better) - Submit images + accurate labels + optional rich metadata (GPS if possible, date/time, crop type, brief damage description, location name) Example Images (placeholders) Healthy / Non-damaged Crop (Label 0): [Placeholder: Insert high-quality example photo of healthy green banana plantation or cacao field in Dominican Republic under sunny conditions] Damaged Crop (Label 1): [Placeholder: Insert high-quality example photo of storm-damaged crops — e.g., flattened banana plants, flooded field, or visibly broken cacao pods/leaves after hurricane] Submission Guidelines - Submit via the official FLock × UNDP hackathon form (link to be provided at kickoff) or DoraHacks.io submission portal. - Preferred format: ZIP package containing images + CSV with columns: image_filename, label (0 or 1), gps_lat, gps_lon, timestamp, crop_type, location_description, notes. - All data must be collected ethically and with landowner/farmer consent where required. - Submissions will be reviewed for quality, diversity, and labeling accuracy. Judging Criteria (Total 100 points) | Criterion | Weight | What We Look For | |----------------------------|--------|------------------| | **Quantity of valid high-quality images** | 40% | Volume + consistency | | **Diversity & coverage** (locations, conditions, crop types) | 30% | Geographic & visual variety | | **Image quality** (clarity, lighting, composition) | 20% | Professional-grade smartphone shots | | **Metadata completeness & labeling accuracy** | 10% | Rich, accurate, traceable data | Winners selected by FLock + UNDP jury (with input from local university partners). Prizes & Bounties (Total Prize Pool: $1,000+) - 1st Place (Top Contributor): $500 USD - 2nd Place: $300 USD - 3rd Place: $200 USD - Top 10 Contributors: Recognition certificates + feature in FLock × UNDP case study + potential continued paid data roles or AI Arena bounties - All Participants: Official participation certificate + chance to be highlighted in press and investor materials Prizes paid via local on/off-ramps or preferred method (fiat/crypto). Timeline (Aligned with Field Trip – Last week of May 2026) - Registration Opens: Immediate (pre-field trip) - Public Kickoff: During FLock × UNDP field trip (last week of May) – in-person launch with government/university partners - Data Collection Window: 3–4 weeks post-kickoff - Submission Deadline: Mid-June 2026 - Winners Announced: Late June 2026 (live or virtual ceremony) Exact dates will be confirmed with university partners during pre-trip ramp. Eligibility - Current students (undergraduate or postgraduate) from Dominican Republic universities and higher-education institutions - Individuals or small teams (up to 5) welcome - Open to all disciplines (agronomy, computer science, environmental science, photography, etc.) Partners & Supporters FLock.io • UNDP Dominican Republic • Local University Partner(s) • AGRODOSA / CONACADO / ADOPEM (field access support)