INDUSTRY FIELD GUIDE
Vineyard cluster and maturity records for AI evaluation
A cluster photograph is useful for counting.
NAICS 111332 · Grape Vineyards
A cluster photograph is useful for counting. A berry sample adds a different kind of evidence, and the two must refer to the same vineyard block and visit before an assistant combines them. University of California guidance on table grape harvesting gives context for maturity and handling. A buyer still needs to specify whether the task is count, sample reconciliation, or visual evaluation. UC harvest and handling guidance for table grapes · UC Kearney vine crops research
Build a fictional block map and three visit records. Match a cluster image and a documented berry measurement for one visit. Give another visit the same row number in a different block. Ask the model to show its matching evidence and flag where a reported reading should not be reused. Keep table and wine grape examples labeled separately if the buyer wants both.
Protect grower identities, block coordinates, winery quality targets, and contracted yields. Use images from an owned row or approved demonstration. Keep the exercise about linking evidence, with the grower retaining harvest decisions.
Original Fiund exercise
A row number repeats across blocks and cannot identify a grape sample alone
A fictional table-grape visit log has original image and berry-sample descriptions. It represents no real maturity result, winery target, crop yield or harvest decision.
| Record | Source | What it records |
|---|---|---|
| VC1 | Block A visit V1 | Day 2, row 3: cluster image I1 and berry sample S1 explicitly linked to V1. Reviewer verifies 4 visible clusters; artificial sample index 16. |
| VC2 | Block B visit V2 | Day 2, row 3: sample S2 index 18. Different block and visit; no I1 relationship. |
| VC3 | Later A visit V3 | Day 5, row 3 image I3, sample field blank. Summary draft reuses S1's day-2 index as a day-5 result. |
Question for the model
Which image and sample can be paired, and can S2 or the earlier S1 reading fill V3's missing measurement?
Show the reference answer and checks
VC1 explicitly pairs I1 with S1 for Block A/V1 on day 2, supporting the stated visible-cluster count and separate sample index. VC2's matching row number belongs to Block B/V2, and S1 belongs to an earlier visit than VC3. Neither supplies a day-5 V3 measurement. Keep visual count and sample result distinct, and request a visit-specific sample rather than carry a prior reading forward as observed maturity.
- Joining records by row number without block and visit identity.
- Reusing an earlier sample as a new visit's measured value.
Fields to preserve
- Block, row and visit
- Defines the collection event.
- Image-to-sample cross-reference
- Supports the pairing.
- Count versus sample metric and date
- Separates different evidence types.
Test another case
Add V3's image-linked sample with its own method label. Require method comparability review before interpreting the numerical difference as a trend.
Use original logs and owned authorized imagery. Exclude coordinates, identities, quality targets and contracted yields; the grower retains harvest decisions.
Discuss a contribution
Describe the sources you control, the task they could support and what has already been checked. Use the existing-data assessment for an archive or the collection brief for a new recording project. The contributor agreement defines Fiund’s licensing authority and compensation. A guide or fictional exercise does not establish an accepted collection or a sale.