This dataset is released for research use. Access is reviewed and granted manually by the maintainers. Please state your name, affiliation, and intended use.

191-counting

1,549 counting questions derived from NEU-DET's human-drawn boxes. This repository is a MIXED container.

The task

One question per image: how many separate surface defect regions are on it? The answer is a single integer — the number of the canon's human-drawn boxes on that record. Regions of different kinds are counted separately; each separate region counts once.

Roles: this is an answer-only tier — annot is the machine-parseable gold AND the training target, in the format the query specifies; reasoning is null by design, because no teacher was called for this repository.

Split, in words

This repository is a MIXED container: it holds records from the canon's train split and from its validation split, so the HF split is named train and the truth for every record is its own metadata.split and metadata.eval_lock.

where the record comes from	records
canon `train`	1,522
canon `validation`	27
total	1,549

load_dataset(...)["train"] therefore hands back train 1,522 + validation 27. A record whose metadata.eval_lock.locked is true is evaluation material wherever it is read; common/overlap.py classifies every record in this repository with 0 unknown, so the container is checkable by the reader rather than promised here.

Overlap / de-duplication (§8)

Cross-family evaluation lock — metadata.eval_lock (manifest revision fe6e286912b0, generated 2026-09-08). Every record of this repository, locked or not, carries metadata.eval_lock, computed by forge_model/common/overlap.py::Overlap.stamp_for against common/overlap_manifest.json at that revision — so within this repository the absence of the key cannot occur. Shape: {"locked": bool, "against": [{"repo": …, "split": …}, …], "own_split": …, "manifest_revision": …, "manifest_generated": …}. locked is true when the image is evaluation material anywhere in the corpus; own_split marks a record locked by its own split. The per-record field is the authority — the count here is quoted once, at this revision, and a later manifest may change it: 27 of 1,549 records (27 distinct images) are locked — by column: 0 by the cross-family manifest, 27 by their own split. In words: this repository is a MIXED container published under the HF split name train; the 27 locked records are the canon's validation records, and a training pool built from AI4Manufacturing/191-counting must exclude them — Overlap.partition / assert_train_pool_clean read the field and will. A stamp whose manifest_revision differs from the current manifest is stale, not wrong — recompute it (Overlap.stamp_is_current); a record with no stamp has not been checked against the corpus as it now is.

What is NOT asked here, and why

The per-class count question (how many <class> regions) lives in the sibling 191-qa repository, not in this one. For one class it is not asked at all: pitted_surface's modal answer is 1 on 0.6346 of the records that carry it, over the 0.6 flag line — a model answering "1" every time would score that much without looking.

class	modal answer	modal share	over 0.6?
`pitted_surface`	1	0.6346	flagged
`rolled-in_scale`	2	0.5367	—
`crazing`	2	0.4833	—
`scratches`	1	0.4333	—
`patches`	2	0.2953	—
`inclusion`	1	0.2906	—

Current setting: PER_CLASS_PITTED = drop_question (drop_question (ii, default)), the knob in forge_model/annotate/191/rung_config.py. The alternatives, recorded so the choice is visible rather than inherited: drop_records (i) · keep_with_flag (b).

Why: Stage 4 measured the per-class count question's modal share at 0.6346 for pitted_surface (flag line 0.60); the total-count distribution was never flagged and is class-agnostic.

Which records are not in this rung

251 of the canon's records are excluded. A record can fail several conditions, so the per-condition tallies overlap; these combinations are disjoint:

excluded by	records
declared_exception	2
declared_exception + instance_set_not_fully_legible + repeated_class_box_pair	1
declared_exception + repeated_class_box_pair	2
instance_set_not_fully_legible	246

Legibility floor

The corpus floor is 16 px rendered min side, measured at the resolution the model sees: 200x200 native renders to 196x196 (scale 0.98), so the 2.36 MP cap never binds here and a 16.3 px native box is the boundary. The unit is per-INSTANCE, not per-record-largest.

A count is answerable only if EVERY region can be seen, so a record enters this rung only when its whole instance set is legible: 247 records are out on that condition alone.

This illegibility is intrinsic to the source: 200 px images cannot be re-rendered larger, so no resolution policy changes it. Sub-floor shares by class are on the canon's own card.

How this repository was checked

A verifier battery reads the canon and the committed query pools and knows nothing about how the builder works, so it and the builder can only agree by both agreeing with the canon. On this rung:

check	result
records	1,549
problems	0
template coverage	191/counting 37/37
planted faults caught	10/10

The gold of every record was re-derived from the canon and required to equal what shipped; every query was required to be its pool's template at its recorded index; every record's eval_lock was recomputed; and no query lets a class phrase govern an inflecting verb (common/agreement_probe). Evidence: reports/191/stage5_rungs.html, reports/191/stage5_catalog.html.

Provenance

Derived from AI4Manufacturing/191 at revision eff72aff931d — the canon's own images and human-drawn boxes, unchanged. No model produced any part of this repository: every answer is a function of the canon's geometry.

what	value
canon revision	`eff72aff931d675718c02635110e8b0d8fb3f1d2`
canon shard sha256 (train)	`5db03a541430a3d4f1fd2342c8461e9a0e2f990085ac5bff31f4fcd52af347f4`
canon shard sha256 (validation)	`e8c2aed498a59e71b48457528044aa31ad8dca1a365a3c3820403b9dc154e62b`
builder	`forge_model/annotate/191/build_counting.py` (`558b4727bcfabf5c`)
salt	`191-rungs-v1/counting`
cost	$0 · 0 hosted-model calls

query pool	templates	`pool_sha16`
`191/counting`	37	`01a3aae9b123fb0b`

Source: NEU-DET (Song & Yan, Applied Surface Science 2013; the NEU surface defect database), packaged by the third-party GitHub project siddhartamukherjee/NEU-DET-Steel-Surface-Defect-Detection (id 361085234, license null). The source tree ships no LICENSE file and the README calls the images NEU's "open surface defect database", so license: other (research use) is the strongest claim this data supports.

Loading

from datasets import load_dataset
ds = load_dataset("AI4Manufacturing/191-counting")["train"]          # a MIXED container — see "Split, in words"
import json
md = json.loads(ds[0]["metadata"])
md["split"], md["eval_lock"]["locked"]              # the truth for THIS record

Access is gated manual.

Downloads last month: -

Total file size:

23.6 MB