How shreyaaa463/soilhealthandmicorbialbiodiversitydashboard works
Project Type & Purpose
• This is a full-stack, AI-powered web application (“SoilSense Dashboard”) for assessing soil health, estimating microbial biodiversity, and providing crop/fertilizer/water recommendations.
• Frontend: Single‐page or multi‐view HTML/CSS/JS (Tailwind, Chart.js).
• Backend: Flask app in Python, serving web routes & APIs.
• ML & Rules: Random Forest model + rule-based logic modules.
• Data: Local CSV files (soil rules, fertilizer info, crop rotation) + external weather API (OpenWeatherMap).
File Structure Overview
• app.py: Flask entry point
• templates/: Jinja2 HTML views
• static/: CSS & client-side assets
• utils/: Core recommendation & diagnosis logic (crop_recommend, fertilizer_recommend, water_requirement, problem_diagnosis, weather fetch)
• Models/: Training & inference (train_model.py, saved model artifacts)
• data/: CSV datasets driving rule engines & training
• evaluation/: Accuracy checks for ML outputs
• requirements.txt: Dependency manifest
Key README Insights
• Architecture is layered: Presentation (HTML/CSS/JS) ↔ Application (Flask + utils) ↔ Data/Models (CSV + ML pipelines) ↔ External Services (Weather API).
• Emphasis on real-time weather integration, AI models, and visualizations.
a. Frontend Layer
• Browser client: renders index.html, pulls CSS/JS from static/styles.css.
• Visualization libs: Chart.js, Matplotlib/Seaborn (server-side image generation or JSON data for client).
b. Backend Layer (Flask)
• HTTP Routes in app.py:
– GET / → serve index.html
– POST /analyze → invoke utils modules + ML inference
• Utils modules:
– problem_diagnosis.py (soil deficiency logic)
– crop_recommend.py (loads Random Forest model + label encoder)
– fertilizer_recommend.py, water_requirement.py, weather.py
c. Data & Model Layer
• Local CSV files (in data/) for rules & training
• ML Model artifacts (Models/class_names.pkl, label_encoder.pkl)
• Training script (Models/train_model.py) – separate pipeline
• Evaluation scripts (evaluation/check_accuracy.py)
d. External Services
• OpenWeatherMap API for live weather data
e. Future (optional)
• Authentication & Firestore (Firebase)
• Satellite data (Google Earth Engine)
Relationships & Interactions
• Client → Flask API: user inputs → POST request
• Flask → utils.weather → External Weather API → JSON response
• Flask → utils modules + Models/*.pkl → generate recommendations
• Flask → templates + Chart.js/Matplotlib → render results back to client
• Evaluation scripts (offline) read model outputs + ground truth from data/
Architectural Patterns & Principles
• Layered separation: Presentation ↔ Business Logic ↔ Data/ML
• Modularization: each recommendation/diagnosis rule lives in its own module
• Separation of training (Models/train_model.py) vs inference (utils/*.py)
• Rule-based + ML hybrid pipeline