README.md

metadata

tags:
  - smart-manufacturing
  - sft
  - industrial
  - vision
license: other
pretty_name: D25
extra_gated_fields:
  Name: text
  Affiliation: text
  Intended use: text
extra_gated_prompt: >-
  This dataset is released for **research use**. Access is reviewed and granted
  **manually** by the maintainers. Please state your name, affiliation, and
  intended use.

D25

Visual anomaly detection & localization (1-class protocol). 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

10,821 records (test=2162 · train=8659). 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)
annot	str	the answer — for this dataset: the plain-text image-level label good or anomalous (VisA is binary — no fine-grained defect types). Pixel-level localization is a separate task whose target is the mask 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]	(D21 only) multi-region masks

Task, mask & split

This dataset supports two levels of the anomaly task:

• Image-level detection — query asks only whether the pictured object is good or anomalous, and annot is the plain-text answer good or anomalous.
• Pixel-level localization / segmentation — for every anomalous image the mask column carries the ground-truth defect mask: a binary image (pixel 1 = defect, 0 = background) at the input resolution. Normal images have no defect and therefore no mask (null). A model addressing the localization task is expected to output a binary mask image of the same height×width (1 = defect pixel, 0 = background); this repo ships that mask as the localization target.

Split — one-class (1cls) protocol. train = normal images only (no anomalies, no masks); test = normal + anomalous images, with a mask on each anomalous image (see the exact counts under Records). This is the standard unsupervised one-class anomaly-detection protocol; VisA's supervised 2cls and few-shot protocols (the same images under a different train/test split) are not included here.

Provenance

Underlying dataset: VisA. Upstream license: CC BY 4.0 (this card is license: other; respect the upstream terms). Converted read-only from the raw source into the unified schema; conversion script: D25/convert_d25.py, published with publish/push_to_hf.py, both in AI4Manufacturing/forge_model.

Overlap / de-duplication (§8)

Subset of MMAD's image pool. Each record carries metadata.image_sha256 so overlapping images can be kept entirely on one side of a train/eval split.