metadata

pretty_name: Clker.com Images
size_categories:
  - 100K<n<1M
task_categories:
  - image-classification
  - zero-shot-image-classification
  - image-to-image
annotations_creators:
  - found
language:
  - en
multilinguality:
  - monolingual
source_datasets:
  - original
configs:
  - config_name: default
    data_files:
      - split: train
        path: clker-images.jsonl.zst
    default: true
tags:
  - image
  - clipart
  - text
license: cc0-1.0

Dataset Card for Clker.com Images

Dataset Summary

This dataset contains 140,313 public domain clipart images collected from Clker.com. Clker.com hosts user-shared vector clip art that is explicitly released into the public domain (CC0). The dataset includes the images themselves along with metadata such as titles and tags associated with each image.

Languages

The dataset is primarily monolingual:

English (en): All image titles and tags are in English.

Dataset Structure

Data Files

The dataset consists of:

Image files stored across multiple ZIP archives (e.g., images_00.zip, images_01.zip, etc.) in the archives/ directory. Each archive contains approximately 1000 image files.
An index file archive_index.csv mapping image filenames to their respective archive files. The CSV format is: Archive,Filename.
Corresponding metadata in a compressed JSONL file: clker-images.jsonl.zst. Each line in this file corresponds to one image.

Data Fields

The clker-images.jsonl.zst file contains the following fields for each entry:

id: An identifier for the image on Clker.com.
title: The title given to the image.
tags: An array of strings representing tags associated with the image.
download_url: The original URL from which the image was downloaded on Clker.com. The filename part of this URL typically corresponds to the Filename listed in archive_index.csv.

The archive_index.csv file contains:

Archive: The name of the ZIP archive file (e.g., images_00.zip).
Filename: The name of the image file contained within the archive.

Data Splits

The dataset contains a single split:

Split	Description	Number of Examples
`train`	Clipart images and metadata	140,313

Total dataset size: 140,313 entries

Additional Information

Licensing Information

All assets in this dataset are derived from Clker.com, which states that all clipart is released under the Creative Commons 0 (CC0) license, dedicating the work to the public domain. Assets may be copied, modified, distributed, performed, or otherwise used for any purpose, including commercial, without asking permission or giving attribution.

Full license text: https://creativecommons.org/publicdomain/zero/1.0/