INDUSTRY FIELD GUIDE
Sawmill data for AI board and grade-mark review
A sawmill can propose a dataset connecting board identity, visible characteristics and the grade information actually recorded.
NAICS 321113 · Sawmills
A sawmill can propose a dataset connecting board identity, visible characteristics and the grade information actually recorded. SPIB explains lumber grade marks. Those marks provide context for a task that keeps declared grade separate from what an image alone establishes. SPIB lumber grade marks
Create original board photographs with neutral identifiers and a fictional tally sheet. Show the same board before and after crosscutting, with one resulting piece losing the visible identifier. Ask the assistant to match only the pieces supported by the cutting record. Add a moisture reading tied to a different board with a similar grain pattern. The reviewer key should reject that association even if the number looks plausible.
Annotate parent board, cut event, child piece, measurement location and source region. Reserve a new species description and camera position for evaluation. Score incorrect piece associations, missing moisture context and unsupported grade claims independently. Include a complete case where the cutting log preserves every child identifier, so the model can demonstrate a supported match as well as identify gaps.
Start with authorized sample boards and self-authored records. Actual customer orders, grading documents and worker images need permission appropriate to their source. A licensing brief should identify whether the buyer needs document association, visual observations or both, then establish the required views and independent review before collecting a larger mill archive.
Discuss this original manufacturing sample with Fiund
Original Fiund exercise
Moisture belongs to the dated measurement
An owned board photograph and a fictional tally compare visual evidence with two recorded spot readings.
| Record | Source | What it records |
|---|---|---|
| S1 | Tally at 08:00 | Board B6 in intake rack; spot reading 19% at point P. |
| S2 | Tally at 15:00 | B6 in sample rack; reading 15% at point P using the same recorded method. |
| S3 | Photo | B6 photographed at 15:05; identifier and a visible knot shown. |
| S4 | Label request | Attach nearest prior measured moisture; grade field intentionally blank. |
Question for the model
What should accompany the photo?
Show the reference answer and checks
Attach the 15% reading recorded at 15:00 at point P for B6, because S2 is the nearest prior reading to S3. Keep the earlier 19% as history. The photo supports a visible-knot annotation, while S4 leaves the grade unresolved.
- Inferring whole-board moisture from the photo or one spot
- Assigning a grade from a knot without grading evidence
Fields to preserve
- reading_time
- Supports as-of linkage
- measurement_point
- Avoids generalizing one reading across a board
- visible_feature
- Keeps imagery separate from grade determination
Test another case
Use a closer-in-time reading on another board and require exact specimen matching before ranking by time.
No authentic grading stamps, supplier tickets, construction advice or acceptance criteria.
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.