INDUSTRY FIELD GUIDE
Wood furniture assembly data for checking hardware fit
Furniture assembly is a good place to test a common AI error.
NAICS 337122 · Nonupholstered Wood Household Furniture Manufacturing
NAICS 337211 · Wood Office Furniture Manufacturing
Furniture assembly is a good place to test a common AI error. Two similar brackets may differ only in handedness or hole spacing. NIST’s assembly benchmarks use controlled tasks such as locating parts and joining them correctly. A furniture maker can turn that idea into a demo with its own unbranded hardware. NIST assembly test methods · NIST real-world manufacturing robotics benchmarks
A correct kit needs the right part and the right revision
Put a mirrored bracket, the correct bracket, and two compatible fasteners beside an owned training panel. Pair each attempt with an invented kit card and panel revision. Record the fit check, the reason a part was rejected, and the first view that made the orientation clear. Add a second panel where the old card looks plausible but calls for the wrong pattern. Score part selection, alignment, and abstention when the board revision is missing.
Keep customer assembly instructions, product designs under license, packaging art, and branded fittings separate. A bench with loose parts is enough to show the decision.
Describe a hardware kit dataset
Original Fiund exercise
A matched hardware bag has the wrong handed leaf
An original paper cabinet kit uses mock hardware bag IDs and separate handed pattern markers.
| Record | Source | What it records |
|---|---|---|
| 1 | Kit rev B | Bag H4 should contain left leaf L with two offset holes and right R with three. |
| 2 | Bag image | H4 label visible; contents show two three-hole leaves. |
| 3 | Drawing note | Hole patterns are distinct under the marked-face view, which the photo shows. |
Question for the model
What discrepancy belongs on the kit's evidence card?
Show the reference answer and checks
The bag ID agrees with record 1, but the contents appear to include two R-pattern leaves and no L-pattern leaf under records 2 and 3. Keep the H4 package match separate from the handed inventory mismatch. The total number of leaves is still two.
- Treating bag ID as contents verification
- Checking two leaves but not one per hand
Fields to preserve
- kit_revision
- Selects expected hardware
- visible_hole_pattern
- Distinguishes handed pieces
- observed_count_by_hand
- Shows duplicate and missing positions
Test another case
Give both hole patterns under opposite face views and resolve the marked face before counting sides.
No genuine hardware specs, furniture CAD, installation guidance or warranty claims.
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.