Dataset Card for Side7

Dataset Summary

This dataset contains artwork collected from side7.com, an online virtual art gallery and community for both novice and veteran artists, as well as art lovers. 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.