Skip to content
AgroVision AI

How it works

What happens when you scan a leaf

AgroVision AI combines a computer-vision model with a reviewed guidance library. Here is each step, including where the system deliberately holds back.

  1. 1. You take a photo

    One tomato leaf, in daylight, filling most of the frame. JPEG, PNG or WebP up to 8 MB.

  2. 2. The photo is checked

    We confirm it is a real, readable image of a reasonable size and not too dark or bright. Location data is removed.

  3. 3. Stored privately

    Your photo goes to private storage. Only you — and authorised reviewers, with every view logged — can open it.

  4. 4. The model analyses it

    A computer-vision model ranks the leaf conditions it was trained on and produces a confidence score.

  5. 5. Confidence check

    If the top score is below a threshold chosen during testing, the result is marked uncertain and no condition is named.

  6. 6. Guidance and history

    You see reviewed guidance for the possible match, cautions and sources. The scan is saved to your history.

Tips for a good photo

  • Use natural daylight; avoid harsh shadows and flash glare.
  • Fill the frame with one leaf and keep it in focus.
  • Make sure the spots, yellowing or curling are clearly visible.
  • Avoid fingers, other leaves or objects covering the symptoms.
  • Photograph the underside too if symptoms appear there.

Reading your result

Likely healthy
The top match was a healthy leaf. Keep monitoring.
Possible disease
The model found a likely match above its confidence threshold. Confirm with an expert before acting.
Uncertain
The model could not match the photo confidently. Retake the photo or ask an extension officer.
Severity
Not assessed. The model does not estimate how serious a problem is.
Start your first scan