Skip to content

Private evaluations for robotics and research teams

Test a model before it guides a field decision.

Bring a crop, a model and a decision your team needs to make. We test privately against reviewed observations, compare performance with an agreed baseline and report failure patterns. Our ten years of raw field data stay in-house.

Information in one observation

Visible plant condition at a recorded time.

Additional context in a verified sequence

The same planter identified across dated visits.

Target format · planter identity verification underway

What a later observation adds to model evaluation

Information in one observation

  • Visible plant condition at a recorded time.
  • Crop and growing context visible in the frame.
  • A care action, when it is recorded alongside the image.
  • A human label or model prediction with its source.

Additional context in a verified sequence

  • The same planter identified across dated visits.
  • Visible changes that can be assessed independently of a single-image prediction.
  • Care actions linked to the interval between observations.
  • A basis for testing whether history improves predictions. That improvement remains to be measured.

Target format · planter identity verification underway

One planter, eight weeks

Inspect the relationship between visits.

This fictional eight-week example illustrates how to link care and observations to the same planter. A later improvement is an observation. Establishing its cause requires a separate study.

  1. Week 1

    aphids

    Action · treated
    First record
  2. Week 2

    aphids, fewer

    Action · treated
    unchanged
  3. Week 3

    no problem found

    Action · pruned
    recovered
  4. Week 4

    no problem found

    Action · watered
    unchanged
  5. Week 5

    leaf scorch

    Action · watered
    worse
  6. Week 6

    no problem found

    Action · harvested
    recovered
  7. Week 7

    no problem found

    Action · harvested
    unchanged
  8. Week 8

    no problem found

    Action · no action
    unchanged

Fictional example · one planter over time

What the private report measures

Growth
Error in estimated growth or change across visits
verified planter identity and comparable dates
Health
Detection of worsening plants, missed problems and false alarms
expert review including unflagged cases
Harvest timing
Error in days and proportion within an agreed harvest window
linked crop, visit and harvest dates

Scope a private evaluation

We confirm feasibility and quote each engagement separately. The evaluation report can be shared according to the engagement terms. Raw records remain inside MicroHabitat.

Protocol
agree on the crop, decision, baseline and success criteria
Customer deliverable
private scorecard, error analysis and coverage limitations
Repeat test
compare a revised model on the same agreed criteria

Website assistant

What would you like to know?

Ask about our crop AI, Farm Bot or field evaluations.

Enter to send · Shift + Enter for a new line

Anthropic receives your question, relevant earlier messages from this conversation and relevant source passages. Privacy