Datasets:
language:
- en
pretty_name: Spawning-15m
license: cdla-permissive-2.0
tags:
- image
Spawning-15m
Summary
Spawning-15m dataset is a collection of about 15 million CC0/PD image-caption pairs for the purpose of training generative image models.
About
Training a state-of-the-art generative image model typically requires vast amounts of images from across the internet. Training with images from across the web introduces several data quality issues: the presence of copyright material, low quality images and captions, violent or nsfw content, PII, decaying dataset quality via broken links, etc. Additionally, downloading from the original image hosts introduces an undue burden to those hosts, impacting services for legitimate users.
The Spawning-15m dataset aims to resolve these issues through collecting only public domain and cc0 licensed images, automated recaptioning of image data, quality and safety filtering, and hosting the images in the dataset on dedicated cloud storage separate from the original image hosts. These innovations make Spawning-15m the largest safe and reliable public image dataset available.
Built and curated with Source.Plus.
Overview
This dataset has two components. The first is the metadata, which contains the image urls, captions, image dimensions, etc. The second component are the images.
Metadata
The metadata is made available through a series of parquet files with the following schema:
id: A unique identifier for the image.url: The URL of the image.s3_key: The S3 file key of the image.caption: A caption for the image.md5_hash: The MD5 hash of the image file.mime_type: The MIME type of the image file.width: The width of the image in pixels.height: The height of the image in pixels.license_type: The URL of the license.
Images
The image files are all hosted in the AWS S3 bucket spawning-15m. The URLs to the images files are all maintained in the metadata files.
Tutorials
License
The dataset is licensed under the CDLA-Permissive-2.0.
Reporting Issues
We've gone through great lengths to ensure the dataset is free from objectionable and infringing content. If you find any issues or have any concerns, please report them to us at [email protected], along with the id of the relevant item.