Final Year Project

Farmer Guardian

Rice leaf disease detection designed for real field conditions.

A lightweight mobile application that helps farmers capture rice leaf images, classify major diseases on-device, read treatment guidance, and use weather, community, and alerts in one practical workflow.

Download for Android android

Authors

Mariyam

MTN-FL-112

Laiba

MTN-FL-122

Supervisor

Mr. Muhammad Umar

Department of Computer Science

Institute

NUML Multan

BS Information Technology

November 2025

Problem

Why the project matters

Manual field scouting is slow and subjective, while many existing AI tools depend on cloud inference or stable connectivity. That leaves smallholder farmers without reliable support at the exact moment they need it most.

Core issue

Farmers need a practical diagnosis tool that works directly in the field, even when internet access is weak or unavailable.

Aim

What Farmer Guardian aims to do

Build a lightweight mobile application that classifies four rice leaf conditions on-device, presents treatment guidance, and extends the workflow with weather, alerts, community, and bilingual usability.

Offline-first scan

Core disease detection remains available without continuous internet.

Farmer-focused flow

Large touch targets and practical support keep the system usable in the field.

Project Deliverables

What the project delivers

A compact mobile system that combines on-device diagnosis with practical farm support modules.

offline_bolt

Offline-first diagnosis

The main rice disease scan runs directly on the device.

biotech

4 rice leaf classes

Healthy, Brown Spot, Hispa, and LeafBlast are supported by the classifier.

partly_cloudy_day

Weather support

Live local conditions help farmers time actions around disease risk.

groups

Community and alerts

Posts, expert help, multilingual support, and admin alerts extend the core workflow.

Technology Stack

How the system is built

React Native Expo TensorFlow.js Google Teachable Machine AsyncStorage Open-Meteo API PHP JSON API

Mobile layer

Native UI, camera/gallery access, navigation, and offline user flow.

AI layer

Google Teachable Machine model, exported as TensorFlow.js, and bundled for local prediction.

Local persistence

AsyncStorage keeps sessions, history, and local app state available offline.

Online services

Weather and community sync stay lightweight and optional.

Architecture Flow

1. Farmer / Admin

Uses the app to sign in, scan, post, and view alerts.

2. Mobile client

Screens handle navigation, image input, and workflows.

3. On-device AI

TensorFlow.js classifies the rice leaf image locally on the phone.

4. Optional online sync

Weather and community data synchronize only when online.

System summary

Farmer Guardian follows an offline-first mobile architecture: the classifier and local persistence live on the device, while weather, posts, and alerts remain lightweight online extensions.

Dataset

H. M. Do, Rice Diseases Image Dataset, Kaggle, 2018.

open_in_new Open dataset reference

Screen purpose

Key points

Select screen

Testing

Validation highlights

Pass

Authentication, AI scan, weather, community, alerts

Android 10+

Physical device validation baseline

AsyncStorage

Local persistence for user state and scan history

Real-time

On-device result generation with confidence output

The core workflows passed under functional testing, interface testing, and real-device inference testing, including registration, image selection, disease prediction, weather retrieval, community sync, and alert broadcasting.

Feature

Expected

Status

Login / Registration

Valid user reaches dashboard with correct role state.

Pass

AI Camera Access

Camera or gallery opens safely on device.

Pass

Disease Prediction

System returns a disease label and confidence score.

Pass

Weather Fetch

Current local conditions are displayed.

Pass

Farmer Network

Posts synchronize and appear on refresh.

Pass

Admin Alerts

Broadcast alerts are visible to users.

Pass

Future Recommendations

Recommended next improvements

dataset

Expand the disease library

Retrain with larger field image datasets and support more rice problems over time.

sync

Queue offline community actions

Store posts and comments locally and synchronize them when the device reconnects.

translate

Extend multilingual coverage

Translate guides, alerts, and interactions more broadly for regional accessibility.

security

Strengthen account security

Add stronger authentication and tighter role protection for broader deployment.

Closing

Thank you

Farmer Guardian was designed as a practical offline-first mobile assistant for rice farmers. Thank you for your time and evaluation.

forum Questions and discussion