Industry field guide
Berry harvest and handling images for AI review
Two photos can show that a berry changed appearance.
- 2022 NAICS
- 111334
- Practice pack
- Example and answer key (JSON)
- Updated
What this guide tests
Which appearance change is supported for A1, and what can be inferred about the cause or B1's earlier condition?
See the specimen and fields · Work through the fictional records
Task and source context
Two photos can show that a berry changed appearance. Establishing when or why it changed requires a handling record. USDA publishes commodity-specific fruit standards, and workplace recording programs have named agriculture among their settings. A berry farm can test how a model handles those evidence boundaries before any buyer requests real handling footage. USDA fruit grade and inspection resources · micro1 workplace recording listing
Use a consented, brand-free training setup and two views of the same sample tray. Label an observed blemish separately from any verified handling event. Create a second tray with a feature that looks similar but has no preceding photo. Grade the assistant on describing the change, matching tray identity, and declining to infer causation without evidence. Maintain separate labels for different fruit and stages.
Exclude pickers’ faces, locations, retailer rejection reports, and commercial packout figures. The farm and any packing partner should confirm authority before real images or handling logs leave their systems.
Industry research
The U.S. Census Bureau’s industry definitions establish which activities these codes cover. The workflow sources in this guide describe the task or record type. Fiund’s pilot below uses fictional material to test a specific decision within that scope.
111334 · Berry (except Strawberry) Farming
This U.S. industry comprises establishments primarily engaged in growing berries.
Build a pilot sample
Set up this pilot
Shoot two marked prop berries on a practice tray, with one face hidden in the first photo. Record observed changes by face and reserve any cause judgment for direct evidence.
Prepare the test packets
- Start with the worked case. Use the fictional record types below as a guide to the source files you need. Give each packet a stable ID and save its source versions with it.
- Change one fact. Make one packet with enough evidence to answer the question and one where the evidence points to a different answer. Keep the other fields fixed so a reviewer can tell what changed.
- Remove the decisive evidence. Give the reviewer a third packet with a plausible near-match in its place. Have them write what can still be established and what evidence would resolve the rest.
Fields the reviewer needs
- Tray, berry and photographed face
- Establishes the comparison target.
- Capture time and visible feature
- Documents the observed change.
- Handling event and causal evidence state
- Separates chronology from explanation.
Use Plan a farm, forest or fishery AI data sample to turn the specimen into a brief with a defined task, source and review owner.
Fictional exercise
A new visible blemish can be documented without proving its cause
Original tray and berry aliases illustrate a staged visual-review task. Text annotations substitute for images here; no picking crew, packout data or real grading decision is represented.
| Record ID | Source | What it records |
|---|---|---|
| BH1 | Tray A before view | Berry A1 has no visible dark patch on the photographed face. Opposite side not shown; tray identity card present. |
| BH2 | Tray A after view and handling log | Same face of A1 has a visible dark patch. Log records a tray transfer between views but no observation establishing how the patch formed. |
| BH3 | Tray B single view | Berry B1 shows a similar patch. No before image or transfer record supplied for B. |
Question for the model. Which appearance change is supported for A1, and what can be inferred about the cause or B1's earlier condition?
Show the reference answer
BH1 and BH2 support a newly visible patch on the matched photographed face of A1. They do not show the opposite side or establish that the logged transfer caused it. BH3 supports a patch observation on B1 at one time, with no evidence of when it appeared. Score identity, observed appearance and causal uncertainty separately rather than treating a temporal sequence or similar-looking patch as proof of handling damage.
How to check the result
Ask for the answer, the record IDs that support it and any fact the evidence leaves unresolved. Have a reviewer compare the response against the reference before deciding whether the sample is ready to expand.
Errors to flag in this case
- Converting a before-after appearance difference into a verified cause.
- Applying one tray's history to another berry with a similar blemish.
Change the inputs and run it again
Add a before view where A1's face is occluded. Require the new-appearance claim to be qualified by the missing visibility.
Score the decision and the evidence separately
Save the reviewer’s expected answer, supporting record IDs and unresolved fields in a separate answer key. Score the model on the decision, its citations and its handling of the deliberate gap. A correct answer citing a different case counts as an evidence error.
Before expanding collection, have a second reviewer run the same checks on the held-out packet. Keep related versions, views and exports of the same case together so they stay on one side of the training and evaluation split.
Check permission and scope
Use owned fruit, cleared staging and original labels. Exclude worker images, locations, rejection reports and commercial packout figures; real collection requires farm and packer authority.
Remove precise farm coordinates and operator identifiers from a review copy. Keep the linkage between lot, sample and outcome intact.
Describe an authorized archive through Fiund’s existing-data assessment, or describe a new capture capability through the collection brief. The contributor agreement covers the terms for participation.
Sources and further reading
Resources for this workflow
Company and program reviews
Use these reviews to compare sourcing, capture or evaluation approaches. Confirm the applicable program, accepted sources and terms with the provider.
Compare the approach
truelabel review for teams collecting physical AI data
Compare how a capture team specifies devices, processing and quality review for an outdoor recording pilot.
Compare the approach
Graviti review for workplace video and multimodal training data
Compare a catalog with commissioned multimodal collection when a particular crop, setting or viewpoint matters.
Related industry examples
Industry guide · NAICS 111333
Harvest imagery for AI from strawberry fields and greenhouses
Useful harvest footage shows why a picker chooses one fruit and leaves another.
Industry guide · NAICS 111339
Stone fruit size and handling data for AI evaluation
A peach can look large in a close-up and still fall into a different size class when measured.
Industry guide · NAICS 111335
Tree nut shell and kernel images for AI inspection
An in-shell image rarely proves what a kernel looks like.