Farmbit

Note de workflow

How RGB imagery can support practical orchard scouting right now

5 min read Note de workflow

A lot of teams talk themselves into waiting. They assume they need multispectral, perfect labels, and a bigger hardware budget before they can do anything useful. Most of the time that is just delay dressed up as rigor. If the job is to narrow the walk, spot change, and come back to see whether the fix worked, RGB already gets you on the field.

En bref

Start with ranking the walk, not with pretending RGB explains everything.

Keep the note modest: where to go, what to inspect, and when to re-check.

Trust grows faster from repeatable checks than from a bigger sensor story.

RGB is good at the parts that usually matter first

RGB will not diagnose everything. It does not need to. In a small orchard, the first win is often much simpler: find the few places that changed, rank them, and stop wasting half a day on the wrong rows.

  • It shows visible change and unevenness well enough to narrow the first walk.
  • It helps compare this week against last week or against the rest of the block.
  • It gives you a practical way to check whether an intervention actually moved the field.

The routine has to stay disciplined or RGB gets blamed for everything

Teams lose trust when they treat every hotspot like a diagnosis. That is not an RGB problem. It is a process problem. The output should be a short action list with a confidence note, not a dramatic claim.

  • Capture on a routine. Consistency beats one perfect flight.
  • Compare against a baseline instead of staring at a single image in isolation.
  • Write the next check in plain language: irrigation, pest pressure, nutrition, or surface condition.

The fastest way to lose a grower is to promise too much

Two false alarms in a row will do more damage than a modest but reliable routine. Keep the promise tight. RGB can help a field team decide where to go first. That is already valuable.

  • Do not market RGB as a clean disease diagnosis engine if that is not what the routine can actually do.
  • Do not ship raw anomaly maps with no next step attached.
  • Do not skip QA notes when blur, coverage gaps, or lighting issues weaken the read.

A useful RGB note is short

The field team does not need another dashboard. They need the next move. If RGB is working, the output should feel boring in the best way: a ranked list, a plain-language reason, and a date for the re-check.

  • Top blocks or zones to scout first.
  • A short why and the first thing to inspect on the ground.
  • A follow-up check a few days later so the team can tell whether the field improved.

Questions

Can RGB replace multispectral?

No. But it can prove the routine much sooner, and for many small orchards that is the real bottleneck.

What if flights are noisy?

Then the tool needs QA gates and confidence notes. A bad flight should weaken the claim, not disappear into the report.

Gardez tout le workflow en vue

Chaque note est plus utile quand on la replace dans la boucle relever, decider, verifier.

Step 1

Survey

Capture on a routine, not on a heroic day

Use a repeatable flight or photo routine so the comparison next week still means something.

Step 2

Analyze

Compare against a baseline

Look for the few places that changed, not a perfect explanation for every pixel.

Step 3

Decide

Rank the walk order

Turn the signal into a short list of blocks or zones worth checking first.

Step 4

Act

Give the field team a real next check

The note should say what to inspect when they arrive, not just that something looks off.

Step 5

Verify

Come back and see if the fix held

The second pass matters. That is where trust grows and false alarms start to fall away.

Ouvrir les notes workflow Voir un exemple de résultat