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).

180

Woven-fabric defect classification (12 defect types; segmentation GT). Category B, task T-B2, in the unified Smart-Manufacturing SFT schema.

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

Records

247 records (train=247). Pixel masks are embedded as a mask image column (multi-region masks).

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: plain-text `{label, defect_type}` — `{good, null}` or `{anomalous, <defect>}`, one of the 12 AITEX defect types (the authoritative AFID code->name map is applied). The binary `mask` is deferred localization GT with seg info in `metadata` — 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. AITEX AFID (Silvestre-Blanes et al. 2019, A Public Fabric Database for Defect Detection) — woven-fabric surface defect detection & segmentation: 106 defect images across 12 defect types over 7 fabrics, each with a binary mask, plus 141 defect-free images.

Query & answer (this repo's SFT task). query is our own instruction template (the dataset ships no question); it names the closed set of 12 defect types and asks for the label + defect type. annot = plain-text {good, null} or {anomalous, <defect>} (one defect type per image). Defect types are named via the authoritative AFID code→name map (from the AITEX afid page — filename code ddd → name): 002 Broken end, 006 Broken yarn, 010 Broken pick, 016 Weft curling, 019 Fuzzyball, 022 Cut selvage, 023 Crease, 025 Warp ball, 027 Knots, 029 Contamination, 030 Nep, 036 Weft crack. The raw defect code and the fabric code are kept in metadata (defect_code, fabric_code).

Mask (deferred localization GT; white = defect area). mask = the first anomalous region, masks = the list of all regions; metadata keeps mask_path/mask_paths + defect_area_fraction (the latter over the union of all regions). A few defect images have two disjoint mask regions (_mask1/_mask2) — masks carries both, mask is the first. One defect image (0100_025_08, Warp ball) ships no mask in the source, so its mask is null and masks is [] (the image-level label is still anomalous) — faithful to the raw data, not fabricated. Defect-free images have mask=null, masks=[]. Localization is deferred.

Split. Single train split (247 = 106 defect + 141 defect-free); AFID ships no official train/test split.

Provenance

Underlying dataset: AITEX-AFID. Upstream license: other (research use; AITEX AFID, Silvestre-Blanes et al. 2019) (this card is license: other; respect the upstream terms). Converted read-only from the raw source into the unified schema; conversion code under 180/ (with publish/push_to_hf.py) in AI4Manufacturing/forge_model.

Overlap / de-duplication (§8)

None notable. Each record carries metadata.image_sha256 so overlapping images can be kept entirely on one side of a train/eval split.

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`	`03e238a92196966a`

Provenance and verification


records	247
carrying a geometry block	247 / 247
instances per record	`0`: 143, `1`: 95, `2`: 8, `3`: 1
total instances	114
image dimensions	4096×256 (246), 3796×256 (1)
`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)	3796×256 → 3808×252, 4096×256 → 4088×252
shipped boxes	114
legible at that render (>=16px there)	30 (26.3%)

⚠ 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 3796×256 is rendered 3808×252 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: 24

Total file size:

176 MB

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