You need to agree to share your contact information to access this dataset
This repository is publicly accessible, but you have to accept the conditions to access its files and content.
This dataset is released for research use. Access is reviewed and granted manually by the maintainers. Please state your name, affiliation, and intended use.
Log in or Sign Up to review the conditions and access this dataset content.
188-region
Region-conditioned defect QA on mobile phone screen — 3653 items, derived deterministically from the
binary segmentation masks of AI4Manufacturing/188. Exact-match gradable → SFT and
RLVR-ready.
Task
"A region is indicated — what defect, if any, is there?" One item per defect instance
(3050 positives: scratch 1372, stain 1120, oil 558) plus 603 clean-region negatives teaching rejection
(every good record + ~half of defective records). The region is conveyed in one of two modes
(metadata.region_mode; overlay 1781 / bbox_text 1872):
overlay— a red rectangular ring drawn on the image around the region.bbox_text— the raw image plus the region as a pixel box[x, y, w, h]in the query text.
Clean boxes sample size AND position from the emitted positive population, so box geometry
separates nothing. Gold = the type name (closed list) or no defect. Verified: zero defect pixels inside any clean box; positive
gold matches the source type on every row.
| field | meaning |
|---|---|
query |
region prompt (overlay or bbox-text) |
image |
photo, or photo with ONE red ring (overlay mode) |
annot |
the type name (closed list) or no defect |
reasoning |
null — deterministic |
cate / task |
B / T-B2 |
metadata |
source, image_sha256, image_wh, region_mode, bbox_xywh, instance_index, gold |
Provenance
Built deterministically (no LLM/teacher; reasoning is null) from AI4Manufacturing/188
— MSD mobile-phone-screen inspection — 1,220 photos, 3 defect types (oil, scratch, stain) + good; each anomalous image has a paired binary pixel segmentation mask (binarized from the palette mask at index>0 (the source encodes defect TYPE in the palette index, so LOCATION comes from the mask and TYPE from metadata)).
Generator: annotate/188/build_188_derived.py in forge_model; machine gates
annotate/188/verify_188.py (all green at build: boxes/gold re-derived byte-identical, clean
regions contain zero defect pixels, mcq gold = the true panel, no leaked-artifact vocabulary, image
set sha256-disjoint from all other cached AI4Manufacturing datasets).
Resolution. Images are native 1920×1080 (not downscaled — preserves thin scratches); all coordinates live in that pixel space (see metadata.image_wh).
Query diversity. query is drawn from a fixed pool of surface variants (paraphrases preserving
the task + answer format), selected by an independent per-record hash; a machine gate confirms no
template correlates with the gold.
The repository name is an internal task code (the source dataset's code is
188).
- Downloads last month
- 6