Hugging Face's logo

Datasets:
hzafar
/
TEXMET

Tasks:
Image Classification
Other
Modalities:
Image
Languages:
English
Size:
10K<n<100K
Tags:
computer-vision
textiles
cultural-heritage
art-history
metropolitan-museum
tapestry
License:
Dataset card Files
xet
 Community
2

Dataset card: align with the JOCCH paper and repo contents

#2
by hzafar - opened
base: refs/heads/main
←
from: refs/pr/2
Discussion Files changed
+105
-59
hzafar
 Owner

Brings the dataset card in line with the published paper (ACM JOCCH, DOI 10.1145/3846195) and with what this repo actually contains. One commit per fix.
Card: rename TeXMET to TEXMET to match the paper73720575
 Card: correct unique object count from 1,697 to 18,6435f6894a0
 Card: directory structure now lists only what this repo contains (300-image sample, no raw folder)ea67d23c
 Card: average file size 1.35 MB -> 1.42 MB (measured 26.49 GB / 18,644 images)69759854
 Card: download success rate 99.274% -> 99.95%, with the paper's collection countse40f9ce6
 Card: curation counts now match the paper (27,360 primary images -> 18,644, 8,716 removed)bcb14e28
 Card: cite the JOCCH paper (DOI 10.1145/3846195, all three authors) instead of the thesis placeholder2d223d32
 Card: add instructions for downloading the full 26.5 GB image set via download.py5f0b26d3
 Card: add Splits section (14,915 / 1,864 / 1,865, seed 42) linking to splits/c2d7cc1c
 Card: state the benchmark purpose (puzzle solving, archaeological textile reconstruction) and add tagsb46cdcc3
 Card: link the JOCCH paper (Just Accepted) and the GitHub repo at the topacb43dfa
 Card: split the paper and code links onto separate lines8298afea
 Card: add department distribution table (14 departments, from the paper)93a697e6
 Card: document the object 21466 / 13562 quirk and how to reconcile download.py output with the splitsdfeafdf7
 splits: fill imageURL for object 21466 additional view (DP307213.jpg) and document download.py reconciliation01c29ee4
 Restore the complete texmet_metadata.json (HF copy was truncated at 765 bytes) and fix the usage example's key350b380c
 Card: drop unverifiable 1.14 aspect ratio (measured 1.21 mean h/w); name the 8% non-RGB images as grayscalec56e3924
 Card: remove marketing language (emoji headers, checkmarks, 'Ready for ML' lines, duplicate quality section)84c0b03b
 Card: update Last Updated date and version history for the September 2026 releasec2e69be1
Publish this branch
This branch is in draft mode, publish it to be able to merge.

Sign up or log in to comment