[Show & Tell] Kanops. Open Access Retail Scenes (v0, 10k+, evaluation-only)

Show and Tell
1

Hi all!

New to the space, but sharing a retail scene dataset for evaluation research from our consulting work where we take images of stores for our analysis and work. We have 1m images on our dataset (added to weekly).

  • HF card: dresserman/kanops-open-access-imagery · Datasets at Hugging Face

  • Folder structure: train/{2014, FullStores, Halloween2024}/Retailer/Subcategory/*.jpeg

  • CSVs: MANIFEST.csv (integrity + EXIF), metadata.csv (relative file paths), checksums.sha256

  • License: evaluation-only; no redistribution or model-weight redistribution

Faces are blurred.

We’re planning weak labels + CVAT tags for v1, over 30 tags in the pipeline.

Feedback welcome on target task definitions and any gaps.

Quick Load:

from datasets import load_dataset
ds = load_dataset(“imagefolder”, data_dir=“hf://datasets/dresserman/kanops-open-access-imagery/train”)

1 Like