On-device AI · Farming in Niger

Everydiseasedleaf
tellsastory.
Yourphonetranslatesit.

Photograph an affected leaf and get the diagnosis and treatment in under two seconds — with no network, anywhere in Niger.

0
diseases detected
0%
works offline
0.0s
per diagnosis
lesion
AgriPredict offline
MilletModerate
Millet downy mildew
96.4%

Recommended treatment available in FR · HA · DJE

The platform

Fromthefieldtothedashboard,
onecontinuouschain.

Mobile, web and backend designed together — for the realities of Nigerien fields, not for a demo.

Offline

On-device AI diagnosis

An INT8-quantised MobileNetV2 runs directly on the phone. 24 disease classes identified in under two seconds — without a single byte of network.

24
classes
3,4 Mo
model
<2s
inference
Secure

Regional SMS alerts

HMAC-SHA256 signed protocol. Alerts reach farmers even without a smartphone.

Android

Offline-first mobile

Room Database, WorkManager sync, SMS OTP auth. Built for areas with no coverage.

Geospatial

Mapping of Niger

Three-level GeoJSON — regions, departments, communes — to read outbreak clusters in real time.

FR · EN · HA · DJE

Four languages

French, English, Hausa, Zarma — the whole agronomic knowledge base, translated.

Web

Expert & admin dashboard

Full RBAC, diagnosis review, model management, push campaigns.

24 diseases

Knowledge base

41 documented treatments, synced to the device for offline reading.

Analytics

Statistics & impact

Metrics by zone, expert validation rate, seasonality of diseases.

The journey

Foursteps,
zerotraining.

From a photo taken at the edge of the field to the SMS alert received across a whole region.

01

Photograph the plant

The farmer frames the affected leaf in the app. No connection is needed — the photo stays on the device.

Capture · 1 tap
02

The AI analyses locally

The TFLite MobileNetV2 INT8 model runs on the phone's processor and compares the lesion against 24 known conditions.

Inference · under 2 s
03

Diagnosis & treatment

The disease is named, the recommended treatment appears, and the sheet stays readable offline indefinitely.

Sheet · offline
04

Review & broadcast

As soon as a network appears, the case reaches the dashboard: an expert reviews it, and the alert goes out by SMS across the region.

Sync · regional SMS

Diagnoses sync automatically as soon as a connection is available — every case can be reviewed by an expert from the web dashboard.

Mobile app

Download AgriPredict

Free. No sign-up to start. Works everywhere in Niger.

Android 8.0+ · v1.0.0 · 32 MB

Agronomic expertise in your pocket

Photograph your crops, get an instant AI diagnosis and treatment recommendations — even without an internet connection.

  • Full offline AI diagnosis
  • Real-time SMS alerts
  • Guidance in French, Hausa and Zarma
  • Works without an internet connection
Scan to download

Open your phone's camera and point it at this QR code.

1Scan the QR code
2Install the APK
3Start diagnosing
Partners pending

Reserved slots for future partners

Official partners will be added once confirmed. For now, these cards stay as neutral placeholders.

Placeholder
Research partner

Reserved for an institute, university, or lab that will support scientific validation.

Scientific co-validation and field trials.

Placeholder
Field partner

Reserved for an NGO, cooperative, or rural support actor.

Awareness, training, and usage feedback.

Placeholder
Institutional partner

Reserved for a public organization or national program to be confirmed.

Coordination, governance, and scale-up.

No partner is shown as official yet. These blocks will be replaced once validated.

Partners to be confirmed
Contact

Let'stalkaboutyourproject

A question, feedback from the field, an interest in collaborating? Write to us.

Support
Niger/Sahel project support
Reply within 48 working hours Your data is never sold

Write to us

Your message goes straight to the project inbox.

Are you affected by, or interested in, the project? Farmer, agronomist, cooperative, researcher, institution or potential partner — write to us here. Any question, field feedback or collaboration proposal is welcome.

0 / 2000

By sending this form, you agree that your name and email will be used solely to reply to you.