🌱 SoilVision — AI-Powered Soil Analysis & Crop Recommendation System

By Team: The Debug Dynasty

24-Hour Hackathon Project (December 2025)

SoilVision is an integrated AI-driven agricultural assistant that analyzes soil images, retrieves real-time weather data, and generates data-driven crop recommendations tailored to Indian farming conditions.

Using Computer Vision, Machine Learning, CLIP embeddings, Random Forest classification, and LLM reasoning, SoilVision empowers farmers to make informed decisions instantly — without the need for laboratory testing.

---

🚜 Why SoilVision?

Indian farmers face several critical challenges:

• ❗ Uncertain crop selection due to soil & climate mismatch
• 🧪 Expensive, time-consuming soil testing
• 🌧️ Lack of real-time weather insights
• 🧭 Regional soil variation (Black, Red, Loamy, Clay, Sandy soils)
• 🌦️ Climate unpredictability and erratic rainfall

SoilVision solves this by transforming a simple soil photo + location into actionable recommendations.

---

⚙️ How It Works

1️⃣ Upload Soil Image

• User uploads a soil photo via the Streamlit UI
• Image is preprocessed using OpenCV (color, texture, morphological features)

2️⃣ Soil Type Classification

• Random Forest Classifier predicts one of 5 soil types:

  • Loamy
  • Clay
  • Sandy
  • Red
  • Black

• Confidence score (0–100%) provided

• CLIP (OpenAI) model assists with cross-modal soil interpretation

3️⃣ Weather Data Integration

• Fetches location-based weather using:

  • OpenWeatherMap OneCall API (3.0)
  • Auto-fallback to 2.5/weather if needed

• Retrieves temperature, humidity, rainfall forecast (7 days), wind, and more

4️⃣ AI Recommendation Engine

The LLM generates a structured response including:

• 🌾 Top 5 suitable crops
• 🧪 Recommended NPK fertilizer ratios
• 💧 Irrigation schedule (rain-aware)
• 🐛 Pest & disease considerations
• 🌱 Season-based advice (Kharif, Rabi, Zaid)

5️⃣ Dashboard Visualization

Streamlit UI displays:

• Soil type + confidence badge
• Weather metrics
• 7-day rainfall forecast (Plotly)
• AI crop recommendations
• Soil image preview
• Pro tips for farmers

---

🏗️ System Architecture

User Input (Image + Location + Season)
                ↓
       Streamlit Frontend (app.py)
                ↓
      Backend Logic (backend_logic.py)
   ┌─────────────────────────────────┐
   │ 1. Image Processing (OpenCV)    │
   │ 2. Random Forest Soil Model     │
   │ 3. Weather API Integration      │
   │ 4. Data Fusion (soil+weather)   │
   │ 5. LLM Recommendation Engine    │
   └─────────────────────────────────┘
                ↓
        Results Dashboard UI

---

📊 Model Performance

Random Forest Soil Classifier

Soil Type	Precision	Recall
Loamy	0.92	0.89
Clay	0.85	0.87
Sandy	0.88	0.86
Red Soil	0.86	0.88
Black Soil	0.89	0.90

• Overall Test Accuracy: 87.6%
• Mean Confidence: 84%

---

💡 Key Features

• ✔️ AI-powered soil classification (CV + ML + CLIP)
• ✔️ Real-time weather integration with smart fallback
• ✔️ LLM-based structured crop recommendations
• ✔️ 7-day rainfall forecasting
• ✔️ Holographic UI theme with glass-morphism
• ✔️ Mobile-friendly interface
• ✔️ Multi-parameter fusion: soil + weather + season
• ✔️ GPS-based location detection

---

🧪 Tech Stack

Frontend

• Streamlit
• Streamlit-Lottie
• Plotly

Backend

• Python
• OpenCV
• scikit-learn
• scikit-image
• Transformers (CLIP model)
• PyTorch

Data & APIs

• OpenWeatherMap API
• Joblib (model serialization)

---

📁 Project Structure

📦 SoilVision
├── app.py                     # Streamlit Frontend
├── backend_logic.py           # Processing + API + ML pipeline
├── soil_rf_model.pkl          # Random Forest soil model
├── requirements.txt
├── gps.html                   # Geolocation utility
├── user_location.json         # Cached coordinates
└── assets/                    # Images, Lottie animations, UI assets

---

🚀 Installation & Setup

1. Create virtual environment

python -m venv soilvision_env
source soilvision_env/bin/activate   # Windows: venv\Scripts\activate

2. Install dependencies

pip install -r requirements.txt

3. Add your API key

Create a .env file:

OPENWEATHER_API_KEY=your_api_key_here

4. Run the application

streamlit run app.py

---

☁️ Deployment

Streamlit Cloud

1. Push code to GitHub
2. Connect repo to Streamlit Cloud
3. Add environment variable OPENWEATHER_API_KEY
4. Deploy ✔️

Docker

FROM python:3.10-slim
WORKDIR /app
COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt
COPY . .
CMD ["streamlit", "run", "app.py"]

---

🔮 Future Roadmap

• 📱 Mobile app (React Native)
• 🌐 Multilingual UI (Hindi, Tamil, Kannada, Telugu)
• 🐛 Pest detection via image classification
• 🧪 Soil pH & nutrient estimation model
• 🌾 Market price integration
• 🚁 Drone imagery support
• 📡 IoT sensor integration (soil moisture, NPK meters)
• 🌤️ Climate resilience scoring

---

🏁 Conclusion

SoilVision bridges the gap between farmers and expert-level soil analysis. By combining computer vision, machine learning, real-time weather APIs, and AI recommendations, the system enables farmers to make smarter crop decisions — instantly.

📈 Impact Potential:

• Improve yields by 15–25%
• Enable 1M+ Indian farmers to make data-driven decisions
• Reduce soil test waiting time from days to seconds

---