D20 / README.md
Y-xvan's picture
D20: MVTec-AD -> T-B1 (unified SFT; viewer-friendly row groups)
69b0409 verified
|
Raw
History Blame Contribute Delete
3.7 kB
metadata
tags:
  - smart-manufacturing
  - sft
  - industrial
  - vision
license: other
pretty_name: D20
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.

D20

One-class anomaly detection with per-category defect naming. 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

5,354 records (test=1725 · train=3629). 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: plain-text {label, defect_type} matching the query form — {good, null} or {anomalous, <defect>} where <defect> is the specific per-category defect name (e.g. {anomalous, broken_large}). The mask column is localization ground truth for a separate, deferred task — 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

What this is. MVTec-AD (Bergmann et al., CVPR 2019) — the standard unsupervised industrial anomaly detection & localization benchmark: 15 categories (10 objects + 5 textures), ~48 fine-grained, per-category defect types, pixel-precise ground-truth masks.

Query & answer (this repo's SFT task). query is our own instruction template (the raw dataset ships no natural-language question — only folder labels + masks). It names the object category, asks the model to decide good vs anomalous, and, if anomalous, name the defect type from that category's own closed set — the valid defect names are enumerated in the query. The answer form is {label, defect_type} — exactly what annot holds ({good, null} / {anomalous, <defect>}). The query does not ask for a pixel mask.

Mask (localization ground truth for a separate, deferred task). Each anomalous image ships a single binary ground-truth mask (mask column; 1 = defect, 0 = background); normal images have mask=null. A text-output model cannot emit a pixel mask directly, so this release keeps the masks as ground truth but does not frame localization as the query task (deferred; masks remain for pixel-level evaluation or segmentation models).

Split. train = normal images only; test = normal + anomalous (all defect types). MVTec-AD ships no separate validation split. Standard unsupervised one-class protocol.

Provenance

Underlying dataset: MVTec-AD. Upstream license: CC BY-NC-SA 4.0 (this card is license: other; respect the upstream terms). Converted read-only from the raw source into the unified schema; conversion script: D20/convert_d20.py, published with publish/push_to_hf.py, both in AI4Manufacturing/forge_model.

Overlap / de-duplication (§8)

Subset of MMAD's image pool; do not put both in train+eval. Each record carries metadata.image_sha256 so overlapping images can be kept entirely on one side of a train/eval split.