Dataset Card for Side7

Dataset Summary

This dataset contains artwork collected from Side7. The dataset includes images along with associated metadata such as titles, descriptions, categories, ratings, and tags.

Languages

The dataset is primarily monolingual:

English (en): Most image descriptions and metadata are in English, though some artists may include multilingual content in their descriptions

Dataset Structure

Data Files

The dataset consists of:

Artwork image files (stored across 8 ZIP archives, labeled images_00.zip through images_07.zip)
Corresponding metadata in JSONL format (side7.jsonl.zst) containing information about the artworks, including IDs, titles, descriptions, and categories
An archive index CSV file mapping image filenames to their respective archive files

Data Fields

id: Unique identifier for the artwork
title: Title of the artwork
description: Description of the artwork provided by the artist
image_url: URL to the artwork image
category: Category of the artwork (e.g., "Fan Art - Anime/Manga")
rating: Content rating of the artwork (e.g., "Everyone")
class: Classification of the artwork (e.g., "Finished Work")
tags: Array of tags associated with the artwork
timestamp: Unix timestamp when the artwork was posted

Data Splits

All artworks and metadata are in a single split with 7,610 entries.