FAO Youth Agrifood Innovation Challenge — Stage III

Diagnose the leaf.
Offline. In her own language.

CropLens is an offline AI crop-disease detection app built for tribal women smallholder farmers across Northeast India — no signal required, no literacy assumed, no delay between symptom and advice.

STAGE III SUBMITTED — AWAITING FAO REVIEW  ·  MobileNetV2 + TFLite  ·  KHASI / GARO AUDIO ADVISORY
The problem

Extension advice rarely reaches the terraced field.

In the hill districts of Meghalaya and across Northeast India, three barriers compound at once.

01

No connectivity

Cloud-based diagnosis tools fail where mobile data doesn't reach — exactly where smallholder farms are.

02

Language & literacy

Text-first apps exclude farmers who read little English or Hindi but speak Khasi or Garo fluently.

03

Delayed diagnosis

By the time a sample reaches an extension officer, the disease has often already spread across the plot.

What it does

Built around the field, not the server.

Every feature answers one constraint a Khasi Hills farmer actually faces.

On-device

Fully offline detection

Runs a MobileNetV2 model converted to TensorFlow Lite directly on-device, so a diagnosis doesn't wait for signal.

Voice-first

Spoken audio advisories

Results and treatment guidance are spoken back aloud (currently in English), removing the literacy barrier entirely.

Fallback

SMS fallback

Where even the app can't run, a lightweight SMS channel keeps the advisory loop alive.

Last mile

Village Champions delivery model

Trained local champions carry the tool into the community, closing the trust gap that pure tech can't.

How it works

From leaf to language, in three steps.

01

Capture

The farmer or a Village Champion photographs the affected leaf using the phone's camera — no upload needed.

02

Detect

The on-device MobileNetV2 / TFLite model classifies the disease locally, entirely offline.

03

Advise

The result is delivered back as a spoken audio advisory — or by SMS where the app itself can't run.

Try it
● LOADING MODEL…

Upload a leaf, see real diagnosis.

This runs the trained MobileNetV2 model directly in your browser — no server, no upload leaves your device.

DRAG & DROP OR TAP
Upload a photo of a crop leaf (JPG/PNG)

Upload a leaf photo to see a sample diagnosis appear here.

CROP · DISEASE
Confidence: —
Voice advisory will play here.
Under the hood

Lightweight by design.

Chosen specifically to run on low-end, offline Android devices.

ModelMobileNetV2Compact CNN backbone for on-device image classification
RuntimeTensorFlow LiteConverted for fast, offline inference on-device
AudioBrowser TTSSpoken advisories, currently in English
ReachSMS fallbackKeeps advisories reaching farmers without the app
Roadmap

Where CropLens goes next.

The current prototype proves the core idea works. These are the planned next steps.

Language Khasi & Garo audio Native-language voice advisories to remove the literacy barrier entirely for local farmers.
Soil GIS soil classification Mapping soil type by location using GIS data to inform more precise crop and treatment guidance.
Advisory Seasonal & regional guidance Crop advisories tailored to the season and specific region, not just the disease detected.
Data Local crop expansion Extending detection to NE India-specific crops beyond the current PlantVillage-trained set.
Progress

FAO Youth Agrifood Innovation Challenge

Stage 1Submitted · May 8, 2026

Concept submission

Initial CropLens concept and application submitted.

Stage 2Submitted · June 30, 2026

Full proposal

Refined proposal document and a separate participant details submission, revised through several passes on word limits, originality, and claim strength.

Stage 3Awaiting results

Final submission

Stage III materials completed and submitted. Result pending from the FAO team.

Team & endorsement

Behind CropLens.

S

Shubham Kumar Jha

Applicant · B.Sc. (Hons.) Agriculture

ICAR RC NEH Region, Umiam — leading the CropLens application.

A

Ayush

Co-applicant

Partnering on the CropLens submission.

K

Dr. Krishnappa Rangnappa

Faculty endorser

Senior Scientist, DCS, ICAR RC NEH — faculty endorsement for the project.

Contact

Follow the project or get in touch.

For collaboration, endorsement, or questions about the FAO submission — reach out directly.

ProjectCropLens
BaseUmiam, Meghalaya, India
StatusStage III · awaiting FAO results

Messages are sent directly to our inbox — no email app needed.