Gaurav6047/Smart-Farmer-AI

🌱 Smart Farmer AI

AI-Powered Crop Recommendation, Plant Disease Detection, Pest Detection, Fruit Classification & Scientific Fertilizer Engine

<p align="center"> <img src="assets/banner.png" width="90%" /> </p>

<p align="center"> <b>Streamlit • TFLite • YOLOv8 • RandomForest • Scientific STCR Engine</b> </p>

⭐ Badges

<p align="center"> <img src="https://img.shields.io/badge/Streamlit-App-brightgreen?style=for-the-badge"> <img src="https://img.shields.io/badge/TFLite-Models-blue?style=for-the-badge"> <img src="https://img.shields.io/badge/YOLOv8-Pest Detection-orange?style=for-the-badge"> <img src="https://img.shields.io/badge/Fertilizer-STCR Engine-red?style=for-the-badge"> <img src="https://img.shields.io/badge/ML-RandomForest-yellow?style=for-the-badge"> </p>

📌 Overview

Smart Farmer AI is a full-stack agricultural intelligence system designed for real-world farmers. It integrates multiple AI models, scientific agriculture datasets, and a rule-based fertilizer engine — all inside one beautiful, mobile-optimized Streamlit UI.

✔ Photo → Disease/Pest/Fruit detection ✔ Soil data → Best Crop → Full Fertilizer Plan ✔ Works offline ✔ Lightweight + Fast ✔ Multilingual (English + Hindi)

🧠 Features

🌿 Plant Disease Detection

• —38-class PlantVillage TFLite model
• —CPU-optimized (5–20 ms inference)
• —High accuracy + mobile-friendly
• —Full confidence score + alerts

🐛 Pest Detection (YOLOv8)

• —Custom-trained YOLOv8 model
• —Real-time detection
• —Bounding boxes + confidence
• —Works for Indian farm pests

🍎 Fruit & Vegetable Classification

• —36-class TFLite classifier
• —Preprocessed for low-power devices
• —High accuracy on common fruits/vegetables

🔀 Auto Image Router

Automatically routes image to:

• —Plant Disease Page
• —Pest Detection Page
• —Fruit Classification Page
• —Or Background Warning

Powered by 64×64 tiny CNN — fast and lightweight.

🌾 Crop Recommendation System (ML Model)

A machine learning system trained using:

• —N, P, K
• —pH
• —Rainfall
• —Temperature
• —Soil Type
• —Region Data

Model Used: RandomForestClassifier

✔ Predicts best crop ✔ Provides confidence score ✔ Uses scaler for normalization ✔ Works offline ✔ Hindi + English support

Files:

models/crop_rf_final.pkl
models/scaler.pkl

🧪 Scientific Fertilizer Recommendation Engine

A professional-grade fertilizer engine built using:

• —STCR equations
• —Indian soil fertility thresholds
• —Micronutrient critical levels
• —Organic nutrient substitution logic
• —Legume rotation credits
• —pH + EC correction rules
• —Commercial NPK-to-bags conversion

Outputs include:

✔ N, P₂O₅, K₂O (kg/ha) ✔ Organic credits deduction ✔ Urea, DAP, MOP bags ✔ Micronutrient recommendations ✔ Soil correction alerts ✔ Agronomic notes ✔ Full calculation breakdown

Uses the following datasets:

models/soil_fertility.json
models/standard_npk.csv
models/stcr_equations.json
models/organic_rules.json

📁 Project Structure

smart-farmer/
│── main.py
│── router.py
│── requirements.txt
│
│── pages/
│   ├── 🌿_Plant_Disease.py
│   ├── 🐛_Pest_Detection.py
│   ├── 🍎_Fruit_Classification.py
│   ├── 🔀_Auto_Routing.py
│   ├── 📊_Crop_Recommendation.py
│   ├── 🧪_Fertilizer_Recommendation.py
│   └── 📘_Fertilizer_Engine_Info.py
│
│── engine/
│   ├── recommender.py
│   ├── stcr.py
│   ├── organic_rules.py
│   ├── brand_converter.py
│   ├── thresholds.py
│   ├── loader.py
│   └── auto_crop.py
│
│── models/
│   ├── plant_disease.tflite
│   ├── fruit_model.tflite
│   ├── pest_model.pt
│   ├── router_model.tflite
│   ├── soil_fertility.json
│   ├── standard_npk.csv
│   ├── stcr_equations.json
│   ├── organic_rules.json
│   ├── crop_rf_final.pkl
│   └── scaler.pkl
│
│── utils/
│   ├── theme.py
│   ├── language.py
│   ├── result_box.py
│   ├── preprocess.py
│   ├── postprocess.py
│   └── model_loader.py
│
└── assets/

⚙️ Installation

pip install -r requirements.txt
streamlit run main.py

☁️ Deploy on Streamlit Cloud

1. 1.Upload to GitHub
2. 2.Go to https://share.streamlit.io
3. 3.Select main.py
4. 4.Deploy → Done 🎉

📱 Mobile-Optimized

✔ Touch-friendly ✔ Responsive grid ✔ Camera input ✔ Smooth animations ✔ Dark/Light friendly

🔮 Future Enhancements

• —AI Voice Assistant (Hindi + English)
• —Offline Android App
• —Weather-aware crop planning
• —Yield prediction model
• —Auto fertilizer schedule based on NDVI

✨ Author

Gaurav — Machine Learning Engineer Building practical & intelligent AI for agriculture 🌱