Skip to content

docs.md

Latest commit

 

History

History
203 lines (152 loc) · 5.77 KB

docs.md

File metadata and controls

203 lines (152 loc) · 5.77 KB

SmartSpray Documentation

1. Overview

SmartSpray is an AI-powered precision agriculture system that detects plant diseases, recommends pesticide use, and controls spraying automatically via IoT.

  • Problem: Farmers often spray pesticides uniformly, wasting chemicals and causing environmental harm. Disease identification in India requires sending images to agriculture departments, causing delays.
  • Solution: SmartSpray uses a camera + ML to detect disease instantly, fetches recommended pesticide schedules, and uses an ESP32-controlled pump + servo to spray only when required. A mobile app lets farmers monitor plant health and manually control spraying.

2. System Architecture

Components

  1. ESP32 (IoT Controller)

    • Controls pump & servo.
    • Receives spray commands from Flask server.
    • Communicates over Wi-Fi.

  2. Webcam (Camera)

    • Captures plant leaf images.
    • Images are sent to Flask server for ML analysis.

  3. Flask Server (Laptop/PC)

    • Hosts ML model for plant disease classification.
    • Manages communication between ESP32 & Flutter app.
    • Stores logs & pesticide schedules in JSON.

  4. ML Model (TensorFlow/Keras)

    • CNN trained on tomato leaf disease dataset.
    • Classifies disease and provides confidence.
    • Maps disease → recommended pesticide schedule.

  5. Flutter Mobile App

    • Interfaces with Flask server via REST API.
    • Displays logs, disease info, spray history.
    • Allows manual override (trigger spray).

3. Hardware Setup

  • ESP32 → Wi-Fi enabled microcontroller.
  • Servo Motors → Rotate to select chemical bottles via tube-switching.
  • Mini DC Pump → Sprays liquid pesticide/water.
  • Relay Module → Switches pump ON/OFF.
  • Power Supply → 5V PSU or Power Bank.
  • Nozzles & Tubes → Deliver controlled spray.
  • Multiple Bottles → Store different pesticide solutions. Servo + tubing select correct bottle.

4. Software Components

4.1 Flask Server (server/app.py)

  • Endpoints:

    • /detect → Receive image, run ML, return disease + spray plan.
    • /command → ESP32 polls to get latest spray command.
    • /override → Flutter app sends manual spray command.
    • /logs → Return spray/disease history.

  • Uses:

    • ml/model.py → Loads trained CNN model (.keras).
    • server/schedule.py → Maps disease to chemical, spray time, servo index.
    • server/logger.py → Saves logs to JSON.

4.2 ML Model (ml/model.py)

  • CNN trained on Tomato Leaf Disease Dataset (Kaggle).

  • Input: RGB leaf image (224×224).

  • Output: disease class + confidence.

  • Classes hardcoded :

    • Tomato_Early_blight
    • Tomato_Late_blight
    • Tomato_Leaf_Mold
    • Tomato_Septoria_leaf_spot
    • Tomato_Yellow_Leaf_Curl_Virus
    • Tomato_healthy

4.3 Schedule (server/schedule.py)

  • Defines pesticide recommendations for each disease:

    {
  "Tomato_Early_Blight": {
    "disease": "Early Blight",
    "spray": true,
    "spray_time": 3,
    "servo_index": 1,
    "chemical": "Copper Oxychloride"
  }
}

  • Returned by /detect.

4.4 Logger (server/logger.py)

  • Stores all detections & actions in logs.json.

  • Example entry:

    {
  "timestamp": "2025-09-03T00:45:00",
  "disease": "Late Blight",
  "confidence": 0.92,
  "spray": true,
  "spray_time": 4,
  "servo_index": 2,
  "chemical": "Chlorothalonil"
}

4.5 Flutter Mobile App (flutter_app/)

Features:

  • Connects to Flask server via REST (http://10.0.2.2:5000 for emulator).

  • Screens:

    • Home → Overview, quick actions.
    • Logs → View detection & spray history.
    • Control → Manual spray override.
    • (Future: Camera capture & upload).

Dependencies:

  • http (API calls)
  • image_picker (optional, for camera uploads)

Permissions (AndroidManifest.xml):

<uses-permission android:name="android.permission.INTERNET"/>
<uses-permission android:name="android.permission.CAMERA"/>
<uses-permission android:name="android.permission.READ_EXTERNAL_STORAGE"/>
<uses-permission android:name="android.permission.WRITE_EXTERNAL_STORAGE"/>

5. Deployment & Testing

  • ESP32 polls /command every few seconds.
  • Activates relay → pump ON.
  • Servo rotates → selects correct bottle.
  • Pump sprays for spray_time seconds.

6. Workflow

  1. Farmer takes picture (via camera or phone).
  2. Flask server runs ML → detects disease.
  3. Flask finds pesticide recommendation from schedule.
  4. Log entry saved (logs.json).
  5. ESP32 polls /command → executes spray.
  6. Farmer monitors via mobile app.

7. Future Improvements

  • Multi-crop disease dataset.
  • Real-time field deployment with drones.
  • Cloud server for multi-farm monitoring.
  • Auto-mixing multiple chemicals.
  • Offline ML inference on ESP32-CAM with TensorFlow Lite.

8. Folder Structure

SmartSpray/
│
├── ml/                  # Machine Learning
│   ├── model.py
│   ├── best_model.keras
│   └── train_model.ipynb
│   └── model.py
│
├── server/              # Flask Server
│   ├── app.py
│   ├── logger.py
│   ├── schedule.py
│   └── gemini_client.py
│
├── flutter_app/         # Flutter Application
│   ├── lib/
│   │   ├── main.dart
│   │   ├── screens/
│   │   │   ├── home_screen.dart
│   │   │   ├── logs_screen.dart
│   │   │   └── control_screen.dart
│   └── android/...
│
├── logs.json            # Detection + spray logs

9. API Endpoints

  • POST /detect → Upload image, return disease & spray recommendation.
  • GET /command → ESP32 polls spray command.
  • POST /override → App sends manual command.
  • GET /logs → Fetch detection & spray history.