CropSCOUT Vision · camera AI
Module
VIS-CAM 01
Runtime
ORT-web · WASM · 1T
Crop
TOMATO
Mode
SQUARE
Frames
0
Infer
— ms
model check pending
Robot home

Camera AI subsystem · live screening

Scout every leaf. Catch disease early.

This is the vision stage of the CropSCOUT rover. The camera feed is read a few times a second, each frame goes through the crop gate, the condition model and four safeguards, and a label is shown only when several frames agree.

Tomato focus: the gate decides leaf, fruit or neither, and the crop is taken as tomato. Switch the crop focus to ALL to let the gate also tell apple, cucumber and lemon apart.

Tomato leaf conditions

  • Healthy
  • Early blight
  • Late blight
  • Septoria
  • Bacterial spot
  • Black spot
  • Target spot
  • Leaf mold
  • Mosaic virus
  • Yellow leaf curl virus
  • Spider mite

Screening support, not a laboratory diagnosis or a spray recommendation. Confirm uncertain or severe symptoms with an agronomist before treating.

01 / input feed

camera · still

Feed the subsystem

Use the live camera, or load one still of a single fruit or leaf in daylight.

Experimental model release. Confirm the crop and leaf/fruit type before acting. Fruit on the plant, yellowing cucumber leaves, whole tomato seedlings and tomato-fruit rots are still misidentified or not covered. This is not a diagnosis or a pesticide recommendation.

No frame loaded Open the live feed or drop one clear crop photo here
On-device inference · frames never uploaded · live feed on phones needs HTTPS · JPEG, PNG or WebP up to 15 MB
Crop or photo type wrong? Override
Crop confirmation

If either automatic result is wrong, select both the known crop and Leaf or Fruit, then run the analysis again.

For condition screening: one close, sharp fruit or leaf filling the square. Covered fruit conditions are screened directly; leaf diseases need a sharp leaf frame.

02 / inference output

on-device

awaiting frame

Output will appear here

Load a tomato, cucumber, apple or lemon frame. Inference runs in this browser; the image is not sent to the server.

  1. 1Frame one leaffill the square · daylight · sharp focus
  2. 2Hold steady3 of the last 4 frames must agree
  3. 3Read the resultcondition · confidence · spray estimate