A super-app that connects farmers directly to buyers with AI-powered crop intelligence.
Built by Team Latent Lords
- Problem & Solution
- Key Features
- App Walkthrough
- System Architecture
- Tech Stack
- Machine Learning
- Crop Insurance (Web3)
- Revenue Model
- Repository Structure
- Getting Started
- Team
Small and mid-sized farmers face low profits and high uncertainty due to poor market access, limited market-price knowledge, inefficient crop planning, and weak decision support. Middlemen exploit this gap, offering low rates and dragging down productivity.
Earthworm eliminates that barrier with a unified platform combining direct farmer-to-buyer trade, AI-powered crop intelligence, learning resources, and real-time support.
| Solution | Description |
|---|---|
| 🛒 Direct Market Access | Farmers sell directly to businesses (B2B) and consumers (B2C), bypassing middlemen. |
| 🤖 AI-Powered Crop Quality | A custom ML model verifies crop quality for transparent purchase decisions. |
| 💬 AI Chatbot Support | Assists farmers and buyers in real time, simplifying the sale process. |
| 🏪 Digital Marketplace | Farmers list crops with detailed info and images for efficient transactions. |
| 📈 Price Trend Estimation | Estimates future market prices of crop commodities for the coming month. |
| 🩺 Disease Detection | AI-powered smartphone detection for fast, accurate crop disease diagnosis. |
| 🧮 Smart Calculators | Crop yield, input cost, soil fertility, irrigation, and pest management calculators. |
| 🌾 Smart Recommendations | Crop scheduling, seed suggestions, crop suggestions, and fertilizer suggestions. |
| 🛡️ Crop Insurance (Web3) | A Web3-based insurance system that cuts down on paperwork. |
- Direct farmer-to-buyer marketplace — no middlemen, fair prices
- AI crop quality scoring on a scale of 10 (freshness, shape & size, texture, color, firmness, batch consistency, damage)
- Crop price prediction using a Random Forest model trained on government commodity datasets
- Crop disease detection — identifies 320+ diseases with remedy suggestions, supports multiple regional languages
- Real-time weather & farm-condition dashboard — temperature, soil moisture, wind speed, rainfall, UV index, and pest risk
- Financial & inventory management — income/expense tracking with running net balance, crop and pricing tracker
- Advanced crop scheduling — tailored sowing, irrigation, pest management, and harvest timelines with reminders
- Online bidding / auction system — single or group bidding with live, timed competition
- Farmer education & engagement — gamified learning system with region-specific videos
- Web3-based crop insurance — instant, low-paperwork disaster claims
Registration & Dashboard — Farmers and buyers register and authenticate, then set up a profile (name, land size, farming method) that seeds all personalized recommendations. The dashboard analyzes live weather and farm-condition data (temperature, soil moisture, wind, rainfall, UV, pest risk) alongside market trends and inventory.
Crop Disease Detection — Farmers upload crop images; an AI model analyzes them for disease symptoms and returns a diagnosis, severity, and remedy — supporting 320+ diseases and multiple regional languages.
Marketplace & Quality Analysis — Farmers list crops with images that are analyzed by AI for a quality score out of 10 (shape & size, batch consistency, color, firmness, texture, damage). Buyers browse, view details, and place orders, with logistics handled through the platform.
Online Bidding — Farmers list crops for single bidding (individual buyers) or group bidding (bulk buyers). Buyers place real-time bids above a starting price until a timer ends, with the highest bidder winning and payment/logistics handled by the platform.
Also included: crop price prediction, financial management, smart farming calculators, advanced crop scheduling with calendar/notifications, and a gamified education hub with region-specific videos.
The app is structured around three primary flows after sign-up: Dashboard (inventory, finance, weather, market trends), Crop Assistance (recommendation, disease detection, scheduling, seed/fertilizer suggestions), and Sell Your Crop (data entry → AI quality rating → bidding/selling).
| Layer | Technology | Notes |
|---|---|---|
| Frontend | Flutter | Cross-platform UI toolkit; smooth performance on Android, iOS, Web, and Desktop |
| Language | Dart | Asynchronous support with async/await; strongly typed with fast JIT & AOT compilation |
| Backend | Firebase | Authentication (Email, Google, etc.) and Firestore for a real-time NoSQL cloud database |
Crop Quality Analysis — A ResNet-50 model fine-tuned on a custom dataset of 1,000 images collected from IPMC markets, manually annotated for freshness, shape & size, texture, color, firmness, batch consistency, and damage. A feedback loop is planned where farmers can contribute labeled data (and earn credits) to continually improve model accuracy.
Crop Price Prediction — A Random Forest regression model trained on an open government dataset of daily commodity prices, producing a 30-day price chart and a one-month price estimate based on current trends.
EarthWorm Insura is a concept Web3-based crop insurance system (built with Solana/Anchor) that protects farmers' harvests, verifies disaster claims via API, and pays out seamlessly to a connected wallet — cutting down on the complex paperwork of traditional crop insurance.
- 1% platform fee charged to buyers
- Filtered advertisements related to agriculture on the platform
- Selling anonymized/aggregated data to government-based platforms (e.g., a national bank of agriculture and rural development)
- A benchmark ~5% fee on farmer bidding transactions
Market: ~140 million farmers in India, ~42% smartphone penetration in rural India (~50 million users). Total addressable market: ₹20,000 crores (Bain & Co AgriTech Report); crop loss from inefficient markets and middlemen: ₹90,000 crores/year (FAO/NABARD). Competitors: Ninja Cart, AgriBazaar, Agrostar.
Earthworm/
├── .vscode/ # Editor configuration
├── crop-prediction-api/ # Backend API for crop price prediction (Random Forest model)
├── project_earthworm/ # Main Flutter application
├── model.ipynb # ML model notebook (crop quality / disease detection experiments)
├── assets/
│ └── screenshots/ # App and pitch-deck screenshots used in this README
└── README.md
These are general setup notes based on the tech stack — adjust paths/commands to match the actual project layout.
- Flutter SDK
- Firebase project with Authentication and Firestore enabled
- Python 3.x (for the crop price prediction API)
cd project_earthworm
flutter pub get
flutter runcd crop-prediction-api
pip install -r requirements.txt
python app.py- Create a Firebase project and enable Authentication (Email/Google) and Firestore.
- Add your
google-services.json(Android) /GoogleService-Info.plist(iOS) to the respective platform folders. - Update
firebase_options.dartwith your project configuration.
Built for a Smart India Hackathon (SIH)-style problem statement on connecting farmers directly with buyers, and augmenting their decision-making with AI.
Earthworm — helping every farmer turn their harvest into a fair, informed sale.