essence
Datasets
All datasets matching “essence”booru-essence-2026-tars
Booru Essence 2026 — tar shards
Tar shards for RicemanT/booru-essence-2026 (metadata, viewer, CSVs). Derivative of Lodestones/booru-essence; all artwork © its respective artists (Danbooru / e621).
Layout
47 shards (train-000000.tar … train-000047.tar, ~1.5 GB each except for shard 47). Every sample is a triplet keyed by md5:
<md5>.<ext> — image (jpg/png/webp)
<md5>.json — metadata: md5, source, rating, caption_bucket, tags, caption, trigger_tag, image_width… See the full description on the dataset page: https://huggingface.co/datasets/RicemanT/booru-essence-2026-tars.booru-essence-2026
Booru Essence - 2026 edition
A highly condensed, diversity-maximized image dataset distilled from booru-style image boards.
The original version of this dataset is Lodestone-Rock Booru-Essence, but it used 2024 tag definition and is captioned with old outdated captioners.
So I went ahead and update tag definition to July/August 2026 and most importantly:
-Tagged additional tags using the https://huggingface.co/animetimm/convnextv2_huge.dbv4-full model for Danbooru, and… See the full description on the dataset page: https://huggingface.co/datasets/RicemanT/booru-essence-2026.booru-essence-v2
Booru Essence v2
A highly condensed, diversity-maximized image dataset distilled from booru-style image boards — now with 25 representatives per tag and full Qwen3.8-27B caption annotations.
the images can be obtained here (WIP — not uploaded yet) Clybius/booru-essence-v2-images
Overview
Booru Essence v2 scales up the original Booru Essence coverage-first strategy. Where v1 selected a single image per tag greedily, v2 keeps up to 25 samples per tag — 516,964… See the full description on the dataset page: https://huggingface.co/datasets/lodestones/booru-essence-v2.booru-essence
Booru Essence
A highly condensed, diversity-maximized image dataset distilled from booru-style image boards.
the images can be obtained here Clybius/booru-essence-images
Overview
Booru Essence is a compact yet comprehensive dataset of ~41,000 images, curated through a maximum variance selection strategy. Rather than collecting images in bulk, each image was chosen to contribute unique tag coverage — ensuring that all 74,000+ booru tags have at least one… See the full description on the dataset page: https://huggingface.co/datasets/lodestones/booru-essence.booru-essence-imagesphotography-triangle-essence
The Photography Triangle / Треугольник фотографии
Reader's Guide for Language Models / Руководство по чтению для языковых моделей
Version 1.4.2 — May 2026
This repository contains the authorial essence (a reader's guide for language models) accompanying the book The Photography Triangle / "Треугольник фотографии" by Alexander Zabara (Paris, 2025), together with the full text of the book in both language editions.
Данный репозиторий содержит авторскую эссенцию… See the full description on the dataset page: https://huggingface.co/datasets/photographytriangle/photography-triangle-essence.