Crop diagnostics at the edge of the field.No connectivity required.

FABLAB AI puts a rugged crop scanner in every farmer’s hand and connects them to a community where they sell their harvest, buy inputs, hire services and learn from each other. Every diagnosis runs on the device, with guidance spoken in the farmer’s own language.

  • 1Sunlight-readable display6.5 in, 1200 nits
  • 2One-button scanWorks through gloves and mud
  • 3Loud front speakerGuidance the whole family hears
Front and back of the FABLAB AI FS-1 handheld scanner, showing a fall armyworm result on the screen and the ring-lit macro camera on the back 123456
FS-1 · 170 × 82 × 17.5 mm · about 320 g
  • 4Ring-lit macro camera50 MP with its own light
  • 5Colour reference stripTrue colour in any light
  • 6Rugged shellIP68, 1.8 m drops
Scan0141 CropMaize ResultFall armyworm92% Time0.8 s NetworkNone
Built on
  • NVIDIA TAOtransfer learning
  • NVIDIA Tritonexpert-review serving
  • NVIDIA CUDAGPU training
  • On-device NPUno cloud needed
12
crops in the launch catalogue
14
languages of spoken guidance
2 days
of scanning on one charge
0
bars of signal needed to diagnose

Figures are target specifications for the FS-1 prototype.

Built for the people who walk the rows

Farmers notice the problem first, so the scanner belongs in their hands. Extension workers and cooperatives see the synced picture and step in where they are needed.

A farmer harvesting rice with a sickle in Karnataka, India
The farmerOwns the scanner and gets an answer at the plant, today.
A farmer harvesting leaves by hand in Indonesia
The extension workerSees synced reports and handles the cases the scanner flags.
Farmers harvesting potatoes in the Dieng highlands of Central Java, Indonesia
The cooperativeSupplies scanners to members and spots outbreaks early.

Photos: Nanditha Gogate (WELL Labs), Arnawakomang and Arief1802 via Wikimedia Commons. Credits

Made for local farmers, from the Mekong to the Andes

There are about 570 million farms in the world, and most of them are smaller than two hectares. They share the same problem: a sick crop, no agronomist nearby and no signal. The FS-1 is built to answer locally, in the local language.

  • 01

    Southeast Asia

    Crops
    Rice · Coffee · Cocoa · Cassava · Banana
    Voices
    Vietnamese, Indonesian, Thai, Tagalog

    Rice blast · 79% · spoken in Vietnamese · offline

  • 02

    South Asia

    Crops
    Rice · Wheat · Potato · Tomato
    Voices
    Hindi, Bengali, English

    Late blight · 90% · spoken in Hindi · offline

  • 03

    Latin America

    Crops
    Maize · Coffee · Soybean · Potato
    Voices
    Spanish, Portuguese

    Coffee leaf rust · 84% · spoken in Portuguese · offline

  • 04

    Sub-Saharan Africa

    Crops
    Maize · Cassava · Cowpea · Cocoa
    Voices
    Swahili, Zulu, Amharic, French, English

    Cassava mosaic disease · 88% · spoken in Swahili · offline

  • 05

    Middle East and North Africa

    Crops
    Wheat · Tomato · Potato
    Voices
    Arabic, French

    Stripe rust · 86% · spoken in Arabic · offline

  • +

    Your region next

    New crop catalogues and voice packs install as downloads, so the same scanner can serve a new country without new hardware.

    Propose a pilot

Regional crop and language sets are launch targets. Each catalogue ships only after it passes accuracy tests on field images from that region. Farm count: FAO.

A crop clinic that fits in a pocket

Everything a scan needs in one rugged handheld: a ring-lit macro camera, a colour reference, a bright screen, a loud speaker, an on-device AI processor and two days of battery.

The back of the FABLAB AI FS-1 scanner, showing the ring-lit macro camera, wide camera, colour strip, scan button and grip
  1. A 50 MP macro camera inside its own ring light, so a leaf is photographed the same way at noon, at dusk or under cloud.

  2. A second camera for whole plants, stems, pods and fruit that are too big for a close-up.

  3. Six printed colour patches the model reads in every clip scan, so it can correct for light before it reads symptoms.

  4. A large, lit button on the side that works through gloves and mud. Speak sits on the other side.

  5. Raised rubber ribs on the back keep the scanner steady in a wet or dusty hand.

  6. Sealed to IP68 and built to survive 1.8 m drops onto hard ground.

1 / 6

Try a scan

Pick a leaf, set the light and focus, then press Scan. Push a slider too far and the scanner refuses the photo instead of guessing.

Sharp Light OK Offline
  1. Image checkSharpness and light
  2. Crop and damageFind the plant and the problem area
  3. Rank problemsCompare against the crop catalogue
  4. GuidancePick curated next steps
Ready when you are Press Scan, or tap S on your keyboard.

Results are a demonstration with sample data. The real scanner runs its models on the device.

A farmer cycling past a ripening rice paddy in Indonesia Photo: CEphoto, Uwe Aranas, CC BY-SA 3.0

No network? Keep scanning.

On small farms around the world, crop problems are found far from good coverage. The model, the guidance library and the voices are stored on the scanner, so the farmer gets an answer in the field, not after a trip to town.

Twelve crops grown on every continent

Each crop gets its own catalogue of diseases, pests, nutrient problems and visible stress, trained and checked one crop at a time.

One scanner per farmer, one picture for the network

Each farmer scans their own crop. When scanners sync, the people who support those farmers see what is spreading and where.

Every farm reports, without a visit

Synced scans arrive with crop, result and date, so the programme sees outbreaks forming and sends officers to the farms that need them most.

Dailyscans from every equipped farm
Flaggedunclear cases go to an officer
A farmer transplanting rice seedlings in Bangladesh

Plan a rollout for your network

It reports likelihood, not certainty

Every result comes with a confidence level and the next most likely options. When the scanner is unsure, it says so and asks for a better image or an expert, instead of guessing.

Below 50%Asks for another image. Offers to send the case for expert review when connected.
50 to 80%Shows the top options side by side and suggests what to check to tell them apart.
Above 80%Gives the likely problem and curated, locally approved next steps.

Put a scanner in every farmer’s hand, and a market at their fingertips

The first FS-1 units go out in 2027, timed to each region’s growing season. We are looking for extension programmes, cooperatives and research partners on every continent to run them.