carimages/carimages-rights

CarImages Rights Manifest

A rights-resolved index of 73,618 Creative Commons and public-domain photographs of cars, published by [carimages.org](https://carimages.org/).

🔗 Browse the archive at carimages.org →

Dataset home · Licensing guide · Attribution index · Photographers · Tools

This dataset contains no images. It is metadata and URLs only, so no ShareAlike obligation propagates to you for using this file. Each photograph remains under the licence recorded in its own row; license_url is authoritative.

Why it exists

Wikimedia Commons holds these photographs but records licensing as wiki templates in inconsistent forms, which makes an ordinary reuse question — which of these may I use commercially without a copyleft obligation? — impractical to answer at scale.

carimages.org resolves that licensing as part of maintaining the archive, and publishes the result here: each licence reduced to three machine-readable obligations.

It also records the photographer, a ready-to-use credit string, SHA-256 and 64-bit perceptual hashes, intrinsic dimensions and a frame-level dominant colour, across 7,038 photographers.

Usage

from datasets import load_dataset
ds = load_dataset("carimages/carimages-rights", split="train")

# everything safe to use commercially with no copyleft obligation
permissive = ds.filter(lambda r: not r["share_alike"] and not r["no_derivatives"])

Where the photographs are

The rows carry Commons URLs, but every photograph is also catalogued on [carimages.org](https://carimages.org/cars/) — organised by make, model and generation, each rendered with its credit block, licence and photographer already resolved. If you want to look at what is in this dataset rather than parse it, start there:

• —Browse by make and generation
• —Browse by photographer — 7,038 contributors
• —How the licensing works
• —Attribution index

Read this before using it

A rights-resolved index of 73,618 Creative Commons and public-domain photographs of cars, sourced from Wikimedia Commons.

This dataset contains no images. It is metadata and URLs only. Because no photograph is redistributed here, no ShareAlike obligation propagates to users of this file.

Why it exists

Wikimedia Commons holds these photographs but records licensing as wiki templates, in inconsistent forms. That makes an ordinary reuse question — "which of these may I use commercially without a copyleft obligation?" — impractical to answer at scale. This file answers it: every photograph's licence is resolved into three booleans.

Distinct photographers credited: 7,038.

Fields

On `carimages_url`: the photograph rendered with its licence, credit and photographer already resolved — useful for eyeballing a row rather than parsing it. URLs ending #p-<id> anchor directly to the photograph; the rest link to the page containing it. Empty for the small number of photographs not yet placed on a published page.

On `dominant_hex`: this is the dominant colour of the entire photograph — including sky, road surface and background — and is not a reliable indicator of the car's paint colour. It is published as a frame-level measurement only.

Sampling frame, honestly

This is not a neutral sample of car photography. It is what one crawler retrieved from a curated list of manufacturer categories on Wikimedia Commons, subject to a fail-closed licence gate. Photographs were rejected for the following reasons, in these quantities:

• —(rejection counts not available in this build)

Notably, files where the source declares attribution is required but records no author are excluded: such a file cannot be lawfully reused by anyone, since compliance is impossible. Their existence is itself a finding about the source corpus.

Contributor concentration is extreme and users should account for it: a small number of photographers account for a large share of the corpus.

Licence

This metadata compilation is released under CC0 1.0 (public domain dedication), which also waives any sui generis database right.

The photographs themselves are not covered by this dedication. Each remains under the licence recorded in its row, and reuse must follow that licence. The license_url field is authoritative.

Provenance and limits

• —Retrieved from Wikimedia Commons between 2026-07-27 and 2026-09-20.
• —Licence resolution is automated. It is believed accurate but is not a legal opinion; source_page_url is the authoritative record in every case.
• —A Creative Commons licence covers the photographer's copyright only. It grants no rights in trademarks, trade dress, or the likeness of any person or property appearing in the image.

Files

• —carimages-rights-2026-09.csv — CSV, one row per photograph
• —carimages-rights-2026-09.jsonl — JSON Lines, same records
• —datapackage.json — Frictionless Data descriptor

Citation

If this is useful in your work, cite the DOI issued by Zenodo for this version.

Citing

Cite the DOI, which is permanently archived and always resolves to the newest version: <https://doi.org/10.5281/zenodo.21683771>

This Hugging Face copy tracks the live archive at carimages.org/data/ and is regenerated regularly, so it may be ahead of the DOI. Cite the DOI for reproducibility; use this copy for the newest data.

Licence

The compilation is released under CC0 1.0, which also waives any database right. The photographs themselves are not covered by that. See licensing and reuse.

Source: <https://carimages.org/data/> · Version 2026-09