CropGuardian — AI Agricultural Threat Detection

Table of contents

1. Executive summary
2. The problem in everyday terms
3. What CropGuardian does
4. Feature 1 — Real-time AI scanner
5. Feature 2 — Mobile-first field experience
6. Feature 3 — Pest wiki
7. Feature 4 — Smart alert history
8. How the AI works (plain English)
9. Why your data stays on the phone
10. How the app is built
11. Getting started
12. Future roadmap
13. Glossary
14. Elevator pitch

1. Executive summary

CropGuardian is a mobile-first web app that helps farmers spot threats early, learn about common pests, and keep a record of what the camera saw — all from a smartphone.

The app combines four tools in one:

1. Live AI scanner — point your camera at the field; the app draws boxes around cattle and other visible threats in real time.
2. Farmer-friendly design — big buttons, bottom navigation, and a dashboard you can read at a glance outdoors.
3. Pest wiki — a searchable guide to pests like locusts, fall armyworms, and aphids, with prevention tips.
4. Alert history — a log of past detections with time and field sector, so you can spot patterns over days or weeks.

Everything runs in the browser. Photos and video frames are not sent to a cloud server for analysis. That keeps the farm private and works better on patchy rural internet.

2. The problem in everyday terms

Farmers lose crops to problems that appear quickly and are easy to miss:

CropGuardian is built for the farmer who already carries a phone and needs something simple, fast, and private — not another complicated dashboard meant for office workers.

3. What CropGuardian does

Think of CropGuardian as four apps in one jacket:

4. Feature 1 — Real-time AI scanner

What you see

When you open the scanner and allow camera access, the app shows your live field view with colored boxes drawn around objects the AI recognizes. Each box includes:

• A label (for example: cow, bird, horse)
• A confidence score — how sure the AI is (shown as a percentage)
• A “Threat Detected” alert when something farm-relevant is found

What the scanner is best at

The scanner uses COCO-SSD, a pre-trained vision model that is strong at spotting large, visible things:

For tiny insects (locusts, aphids), the live camera model is limited — those pests are better covered in the Pest Wiki (Section 6). Future versions plan custom pest models (see Roadmap).

Privacy-first by design

The scanner never uploads your frames to a server for analysis. The neural network runs inside your browser using TensorFlow.js.

5. Feature 2 — Mobile-first field experience

CropGuardian is designed for one hand, bright sun, and muddy pockets — not for a desktop office.

Design choices

You open the site in Chrome or Safari on your phone — no app store download required today (PWA “install to home screen” is on the roadmap).

6. Feature 3 — Pest wiki

Not every threat shows up as a big object on camera. Some pests are small but devastating. The Pest Wiki is CropGuardian’s built-in field encyclopedia.

What’s inside

• Searchable pest directory — find pests by name or browse categories
• Expert-style entries for common threats, including:
  • Locusts
  • Fall armyworms
  • Aphids
• Identification guides — what to look for on leaves, stems, or soil
• Threat levels — how urgent the problem is
• Prevention methods — both organic and standard treatment options

The wiki turns CropGuardian from “camera only” into a complete assistant — detect what you can see live, and learn what you cannot.

7. Feature 4 — Smart alert history

Every time the scanner flags a threat, CropGuardian can log the event to build a farm audit trail.

Over time, this helps answer questions like:

• “Do cows always enter from the east gate at night?”
• “Are birds worse after harvest?”
• “Did our new fence reduce alerts?”

8. How the AI works (plain English)

Neural network

A neural network is software trained on thousands of example images. It learns patterns — “this shape and texture usually means cow” — without a human writing every rule by hand.

COCO-SSD is a ready-made detector trained on the COCO dataset (everyday photos with 80 object types). CropGuardian loads this model in the browser and runs it on each camera frame.

Object detection vs. classification

CropGuardian uses detection because a field may have multiple threats at once, and the farmer needs to know where to look.

9. Why your data stays on the phone

Field images can reveal crop type, farm layout, and location. Keeping inference on-device is a deliberate privacy choice for the farming community.

10. How the app is built

License: MIT — free to use, modify, and share (see LICENSE in the repository).

11. Getting started

For developers or anyone setting up the project locally:

git clone <your-repo-url>
cd Cattle-Detection-Project
npm install
npm run dev

Then open http://localhost:3000 on your phone or computer browser.

Tip for field testing: Use your phone on the same Wi‑Fi as your computer, and visit your computer’s local network address (e.g. http://192.168.x.x:3000) to test the camera outdoors.

12. Future roadmap

CropGuardian is actively evolving. Planned improvements:

13. Glossary

14. Elevator pitch

CropGuardian is a mobile-first agricultural assistant that puts live threat detection, a pest encyclopedia, and an alert history log in every farmer’s pocket. Using TensorFlow.js and COCO-SSD, it highlights cattle, birds, and other visible threats through the phone camera — with confidence scores and instant “Threat Detected” alerts — while keeping all image processing on the device. Combined with a searchable wiki for locusts, armyworms, and aphids, it helps farmers see, learn, and remember what happened in the field. Built with Next.js 15 for the farming community, under the MIT license.

Developed with care for the farming community.