Skip to content

Why we built Crop Help

Small farms, classrooms, and community gardens deserve the same quality plant-health tools as big operations. Crop Help brings early detection and clear, actionable insights to growers who don’t have a data team — simple to use, accurate in the field, and respectful of how you work.

Meet the team Talk to us

Trusted by programs like

Built with and for real programs — from school gardens to community plots and small farms.

Mission and values

How we build and how you use Crop Help

Clear paths for different groups. Each pillar links to a deeper page with details, examples, and next steps.

Impact at a glance

Last updated

  • 0
    Plots monitored
  • 0
    Early detections
  • 0
    Average turnaround
  • 0
    Image batches processed
  • 0
    Gardens supported
  • 0
    Model updates shipped
How we calculate these
Plots monitored

What Distinct production plots or beds with at least one valid scan in the period.

How A plot is counted once per period using a stable plot id. We exclude test environments and deleted records.

Computation rule
count distinct plot_id where scan_valid is true within window
Early detections

What Model flagged issues that were confirmed before visible loss or spread according to grower notes.

How A detection is marked early when the first model flag arrives at least one task or two days before the first manual note of loss or spread.

Computation rule
count detection_id where first_model_ts ≤ first_task_ts OR first_model_ts ≤ first_loss_note_ts − 48h
Average turnaround

What Median time from first upload in a batch to an alert or task creation.

How We compute the median per day across batches with at least one alert. We remove outliers above the ninety ninth percentile.

Computation rule
median(alert_ts − batch_start_ts) winsorized at P99
Image batches processed

What Grouped uploads from a user session or device run.

How Frames taken within ten minutes on the same device join into one batch. Failed uploads are excluded.

Computation rule
count batch_id where status is success
Gardens supported

What Active gardens that created at least one task or issue in the period.

How We use garden id with activity signals and exclude sandbox tenants.

Computation rule
count distinct garden_id where activity_signal is true
Model updates shipped

What Versioned model releases deployed to any tier of users.

How Includes minor and patch versions. Hotfix only releases are included when they change model outputs.

Computation rule
count release_id where touches_inference is true
Data window and sources
  • Window rolling ninety days unless a long range chart is shown.
  • Scope production tenants only. We exclude sandbox and internal testing spaces.
  • Join keys garden id, plot id, device id, and batch id. We use time bucketing at minute resolution to group frames.
  • Quality outliers above the ninety ninth percentile are winsorized for time based metrics. Failed uploads are removed from batch counts.
  • Reproducibility each release writes a materials file with version pins for model, rules, and thresholds.

Grow with confidence

Bring early detection to your garden

Spot issues before they spread. Crop Help makes plant health insights fast, clear, and actionable.

Get started Learn more
  • Built for small growers and classrooms
  • You own your data