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

Roles

Roles: canon repo — annot is the source label, kept machine-parseable as the gold for verification and reward parsing; there is no reasoning column and this repo is not itself a training view. Derived repos (-annotated, -grounding, -region, -mcq) each state their own regime on their own card. Geometry for every record lives in metadata.geometry (below).

190

SEM nanofibrous-material unsupervised anomaly detection (segmentation GT). Category B, task T-B1, in the unified Smart-Manufacturing SFT schema.

The repository name is an internal task code. See Provenance below for the underlying dataset.

Records

45 records (test=40 · train=5). Pixel masks are embedded as a mask image column.

Unified SFT schema

field	type	meaning
`query`	str	the question / instruction (model input)
`image`	Image	the input image (bytes embedded); for multi-image rows, a preview of the first view
`images`	list[Image]	(multi-image rows) all input views / modalities for the row, bytes embedded
`annot`	str	the answer — for this dataset: the plain-text image-level label `good` or `anomalous` (binary; the label is derived from the pixel mask). The binary segmentation `mask` is deferred localization GT, with seg info (`mask_path`, `defect_area_fraction`) in `metadata`; the grayscale `.tif` sources are re-encoded to `.png` for the image column — see Task, mask & split below
`reasoning`	null	no native CoT in these datasets
`cate`	"B"	SFT category
`task`	"T-xx"	unified task id
`metadata`	str (JSON)	split, provenance, `image_path`, `image_sha256` (dedup key)
`mask`	Image \| null	(T-B1/T-B2 only) the pixel ground-truth mask, bytes embedded
`masks`	list[Image]	(multi-image T-B1 / D21) per-view masks aligned with `images` (None where a view has no defect), or multi-region masks

Task, mask & split

What this is. NanoTWICE (Carrera et al., IEEE TII 2017) — SEM (scanning-electron-microscope) images of nanofibrous filter material for unsupervised surface anomaly detection & localization. A handful of defect-free images are used for training; defective images carry pixel-level ground-truth masks.

Query & answer (this repo's SFT task). query is our own instruction template (the dataset ships no question). It asks only whether the material surface is good or anomalous; annot is the plain-text label good or anomalous, derived from the mask (any defect pixel → anomalous). The query does not ask for a pixel mask.

Mask (deferred localization GT). Each anomalous image ships a binary ground-truth mask (mask column; 1 = defect, 0 = background), with mask_path + defect_area_fraction in metadata; normal images have mask=null. Localization is deferred (a text model cannot emit a mask). The source images are 8-bit grayscale .tif; they are re-encoded to .png for the image column so the dataset viewer renders them.

Split. train = 5 defect-free images; test = 40 anomalous images (the standard NanoTWICE unsupervised protocol — all normal images are used for training).

Query text — pooled paraphrases (v2)

Every record's query is drawn from common/vision_query_pools.json[F3/verdict_word], a pool of 40 gate-verified paraphrases of the shipped wording, assigned by a stable hash of the source image path and recorded as metadata.query_template (11 templates in use, top share 17.8%).

The opening sentence is drawn separately (metadata.query_role, a 10-way hand-written pool _role/examining): the shipped sentence "You are an industrial visual-inspection model examining ." is index 0, index 1 is the noun phrase alone as a caption, the rest vary the frame around it — the noun phrase naming the input is this repository's own, verbatim, on every record. Role and ask are hashed independently.

Template 1 is v1's wording byte for byte (0 records keep it); the pass asserted that on every record before rewriting anything.

Template ↔ gold independence on this build: 45 records, 11 templates, worst template p = 0.0636, alpha 9.1e-04, 0 flagged; 10 roles, worst role p = 0.178, 0 flagged → PASS.

Frame-size floor (common/lazy_floors.py, the standing (width, height)-only row): vacuous — the scored split holds a single class, nothing to predict (2 distinct frame sizes).

Answers, images, masks, split and every other field are byte-identical to v1: this revision was issued from the published parquet itself (tools/requery_published.py), not rebuilt from source, and the pixel-identity guard ran on the embedded images (§8 below).

Provenance

Underlying dataset: NanoTWICE. Upstream license: other (research use; Carrera et al., IEEE TII 2017) (this card is license: other; respect the upstream terms). Converted read-only from the raw source into the unified schema; conversion code under 190/ (with publish/push_to_hf.py) in AI4Manufacturing/forge_model.

Overlap / de-duplication (§8)

None notable.

Two identities, and they answer different questions. metadata.image_sha256 hashes the file bytes: it finds byte-identical copies and is blind to a re-encode. metadata.pixel_sha256 hashes the decoded image (mode | size | pixels): it finds the same photograph saved twice. Only the second one settles whether an image is duplicated.

Measured at build time, not asserted afterwards — a violation aborts the build and names the offending records:


images checked	45
distinct by decoded pixels	45
images carrying more than one record	0
images on both sides of the split	0

Geometry (`metadata.geometry`)

Every record carries a geometry block inside the existing metadata JSON string, so that its gold can be re-derived at any render size. No schema column changed; existing loaders are unaffected.

Coordinates are native pixels of the image in that record (coords_frame: "record_image"). scale is 1.0 throughout — this repo publishes at source resolution, nothing was downscaled at publish time.

"geometry": {
  "image_wh":  [W, H],        // dims of the image in THIS record
  "source_wh": [W, H],        // dims of the original source image
  "scale": 1.0,               // image_wh / source_wh; < 1.0 would disclose a publish-time downscale
  "n_instances": 2,
  "instances": [
    { "instance_id": 1, "bbox_xywh": [x, y, w, h], "min_side_px": 65, "class": null }
  ],
  "n_dropped_subminimum": 0,  // components removed by the filters below
  "union_box_fallback": false,// true => boxes are per-class unions, NOT real instances
  "conventions": { ... }      // see table
}

instances is present even when empty. [] means the record genuinely has no defects; an absent block would mean geometry could not be recovered. Those are different states and are never conflated.

Conventions used to derive it

There is no universal definition of "one defect instance" — it depends on the mask the source shipped. This repo's is stated, not implied:

field	value
`algorithm`	`dilate_cc`
`binarisation`	`gt:0`
`connectivity`	`4`
`merge`	`mask_dilate:1pct`
`min_area_px`	`15`
`max_instances`	`None`
`artifact`	`fine`
`fill_floor`	`None`
`legibility_floor_px`	`None`
`min_side_floor_px`	`None`
`spec_sha`	`eb48584beb6f9599`

Provenance and verification


records	45
carrying a geometry block	45 / 45
instances per record	`0`: 5, `2`: 7, `3`: 4, `4`: 9, `5+`: 20
total instances	517
image dimensions	1024×696 (40), 1024×700 (5)
`scale` values present	[1.0]

Computed from this repo's own masks and verified against this repo's own published answers before it was written — a recomputation that disagreed with the shipped gold would have aborted the update rather than overwritten it.

⚠ The 16px floor applies at the RENDER, not at native

min_side_px is in native pixels. The model does not see native: Qwen2-VL caps by megapixels AND snaps each dimension to a multiple of 28. So min_side_px >= 16 is the floor tested in the wrong frame. Measured on this repo:


native → rendered (qwen2_vl @ 2.36MP)	1024×696 → 1036×700, 1024×700 → 1036×700
shipped boxes	517
legible at that render (>=16px there)	400 (77.4%)

⚠ An earlier version of this section reported the inverse — boxes clearing 16px at native and failing at the render — and that number was misleading. It is frame-relative: publishing at a larger native size lets more boxes clear 16 in the published frame, so more can "fail", which penalises exactly the choice that helps. Measured on 179: publishing native (3024) means a box needs >=32px native to be legible at the render and 86.7% qualify; the previous 1024 publish needed >=47px native and only 69.5% qualified. The native republish improved rendered legibility by 17 points while the old metric scored it as 12.5% "broken". The figure above is the comparable one.

Nothing in the data is frame-dependent — geometry is native and complete. Use forge_model/common/adapt_engine.py, which applies the floor at whatever size the consumer renders.

Using it

Coordinates only stay correct if they are rescaled with the image. A patch-based VLM does not render at native size: Qwen2-VL's processor snaps both dimensions to a multiple of 28, so this repo's 1024×696 is rendered 1036×700 and native-pixel boxes are then wrong by a few pixels. forge_model/common/adapt_engine.py regenerates coordinates for a target render size, re-derives counts, and drops records whose gold no longer holds there.

Downloads last month: 66

Total file size:

23.4 MB

You need to agree to share your contact information to access this dataset