README.md

metadata

dataset_info:
  features:
    - name: id
      dtype: string
    - name: image
      dtype: image
    - name: text
      dtype: string
  splits:
    - name: train
      num_bytes: 1192534908282.634
      num_examples: 2905671
  download_size: 210413447679
  dataset_size: 1192534908282.634

E621 Rising Image Dataset v1

2,905,671 images (~1.1TB) with tags.

This is a raw, uncurated, and largely unprocessed dataset. You likely want to use the curated version, available here.

Image Processing

• Only jpg and png images were considered
• Image width and height have been clamped to (0, 4096]px; larger images have been resized to meet the limit
• Alpha channels have been removed
• All images have been converted to jpg format
• All images have been converted to TrueColor RGB

Tags

For a comprehensive list of tags and counts, see here.

Changes From E621

• Symbols have been prefixed with symbol:, e.g. symbol:<3
• Aspect ratio has been prefixed with aspect_ratio:, e.g. aspect_ratio:16_9
• All categories except general have been prefixed with the category name, e.g. artist:somename. The categories are:
  • artist
  • copyright
  • character
  • species
  • invalid
  • meta
  • lore

Additional Tags

• Image rating
  • rating:explicit
  • rating:questionable
  • rating:safe
• Image score
  • score:above_250
  • score:above_500
  • score:above_1000
  • score:above_1500
  • score:above_2000
  • score:below_250
  • score:below_100
  • score:below_0
• Image favorites
  • favorites:above_4000
  • favorites:above_3000
  • favorites:above_2000
  • favorites:above_1000
  • favorites:below_1000
  • favorites:below_500
  • favorites:below_250
  • favorites:below_100