model v4.2 · live · 214 diseases
Show me the leaf. I’ll tell you what’s wrong.
Photograph a sick plant with the phone already in your pocket. You get the disease name, how fast it spreads, and exactly what to buy — in under ten seconds, on 2G, in the language you speak.
- 1.2M
- leaves, photographed here
- 214
- diseases named
- 43kB
- average request
- ₦0
- to use
Covered crops: cassava, maize, tomato, cowpea and plantain.
Diagnose a real leaf.
Pick one of the five sick leaves below and watch the whole thing run — the same four steps that happen on your phone. Nothing is hidden.
no sign-up · no app install
works in the browser
Pick a leaf above
Or drag a photo of your own into this box.
- ·Reading your photo
- ·Finding the leaf
- ·Comparing to 1.2M local leaves
- ·Naming the disease
What you'll get back
- The name of the disease, and what causes it
- How serious it is and how many days you have
- Exactly what to do, step by step, with the price
- An honest score — and a human if the score is low
Sample photographs stand in for the field. Diagnosis text, thresholds and treatment costs are the field guidance for each disease.
Answer returned in 9.4 seconds
43 kB up · 2 kB down · ₦0.31 of data
Built for the phone you already have.
Most farm software assumes a new phone and cheap data. That assumption is why it never reaches the field. We sized every request against a metered prepaid bundle on a 2G signal, and the compression happens on the phone before anything is sent.
data used per answer
- One diagnosis here43 kB
- Typical crop app screen2.1 MB
- Video call to an agronomist18 MB
A diagnosis costs a third of a naira in data. A ₦100 bundle diagnoses 320 plants.
- Nothing to install
- It runs in the browser you already have. No store, no update, no 40 MB download before you can ask a question.
- A phone from 2015
- 1 GB of RAM is enough. The page is 43 kB and the heavy work happens on our side, not on your handset.
- Compressed before it leaves
- Your photo is shrunk on the phone first, so a bad signal costs you seconds, not minutes.
- No account
- No email, no password, no data collected about you. Your photo is deleted after the diagnosis.
Trained on our fields, not somebody else’s.
Nearly every open crop-disease model is trained on single leaves photographed against a white sheet in a lab, mostly in North America and Europe. Put that model in a Nigerian field and it collapses: different cultivars, harsher light, red soil in the frame, two diseases on the same plant at once.
So we collected our own. 1.2 million leaves, photographed on the phones farmers actually own, in the fields they actually farm, labelled with agronomists in each state.
accuracy on 8,400 held-out photos taken in real fields
- Model trained on public foreign datasets41.2%
PlantVillage-style lab photos, plain backgrounds
- Same architecture, trained on our field imagery92.6%
real backgrounds, real light, real co-infection
1,240
A100-hours per full retrain
6 weeks
between retrains, tracking the season
214
disease classes across 5 crops
8,400
held-out field photos in the test set
37
states covered
610
trained collectors
11
languages labelled
Collectors are paid per verified photograph and keep contributing through the season, so the model sees each disease at every stage rather than only at its textbook peak.
What it knows today.
Five crops, 214 diseases, all of it live. If your crop is not on this list the model says so rather than guess at it.
accuracy measured on
held-out field photographs
| crop | diseases | most common | accuracy |
|---|---|---|---|
 Cassavaege · rogo · akpu | 47 | Mosaic disease · Brown streak · Bacterial blight · Anthracnose | 94.1% |
 Maizeagbado · masara · ọka | 52 | Northern leaf blight · Streak virus · Grey leaf spot · Rust | 92.8% |
 Tomatokamaton · tumatis | 61 | Late blight · Early blight · Bacterial wilt · Leaf curl | 93.4% |
 Cowpeaewa · wake · agwa | 29 | Aphid damage · Brown blotch · Mosaic virus · Scab | 89.6% |
 Plantainogede · ayaba · ojoko | 25 | Black sigatoka · Panama wilt · Bunchy top · Cordana spot | 91.2% |
Next in the index
These four are in collection now. Photograph one today and the model says so and hands you straight to an agronomist — it never invents an answer.
- Rice310k leaves collected · ships this dry season
- Yam96k leaves collected · ships this dry season
- Cocoacollection opens November 2026
- Peppercollection opens November 2026
The plant is in front of you. Ask it.
Free for every farmer, in every season, with no account and no data sold. Send one photograph and see what comes back.
open on any phone browser · no sign-up · nothing to install
Running a co-operative or an extension programme?
Bulk enrolment, shared dashboards for your officers, and diagnosis summaries by ward. We train your collectors and the imagery stays yours.
Talk to the field team