datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
zendo-synthetic-data
Zendo Synthetic Visual Reasoning Dataset
Synthetic Zendo-style scenes with associated rules and per-scene tensor
representations. Each scene either follows ("positive", label=1) or violates
("negative", label=0) a rule that is given in natural language and as a Prolog
query.
Splits
split
scenes
train
56475
test
3344
rules total
3439
Layout
images/<split>/<batch>/<rule_id>/<scene_id>.png — rendered scene… See the full description on the dataset page: https://huggingface.co/datasets/sophia1ch/zendo-synthetic-data.british-library-book-images
British Library Book Images
1,080,814 images cut out of 49,455 digitised books (65,227 volumes, ~25 million pages) published
between c. 1510 and c. 1900, digitised by the British Library in partnership
with Microsoft and released by British Library Labs
on Flickr Commons as the "1 Million Images from Scanned Books" release. The books cover geography,
philosophy, history, poetry and literature, in several languages.
The four image types
British Library Labs… See the full description on the dataset page: https://huggingface.co/datasets/biglam/british-library-book-images.amazon-berkeley-objects
Amazon Berkeley Objects (ABO)
A Hugging Face packaging of the Amazon Berkeley Objects (ABO) dataset. The
data content is the official CC BY 4.0 release from
https://amazon-berkeley-objects.s3.amazonaws.com/index.html. This mirror
changes only the packaging: files are grouped into typed Parquet shards, and
every original media file is preserved byte-for-byte and never transcoded.
Images use the datasets Image() feature, 3D product models use the native
Mesh() feature (original… See the full description on the dataset page: https://huggingface.co/datasets/suvadityamuk/amazon-berkeley-objects.openbrush
OpenBrush-75K
A curated dataset of 75,313 public domain artworks with rich, structured VLM-generated captions designed for training image generation models, fine-tuning vision-language models, and art analysis research.
Dataset Description
OpenBrush-75K contains high-quality reproductions of paintings from the Western art canon, spanning from the Renaissance to the early 20th century. Each image is paired with a detailed structured caption generated by a… See the full description on the dataset page: https://huggingface.co/datasets/jaddai/openbrush.british-library-book-images
British Library Book Images
1,080,814 images cut out of 49,455 digitised books (65,227 volumes, ~25 million pages) published
between c. 1510 and c. 1900, digitised by the British Library in partnership
with Microsoft and released by British Library Labs
on Flickr Commons as the "1 Million Images from Scanned Books" release. The books cover geography,
philosophy, history, poetry and literature, in several languages.
The four image types
British Library Labs… See the full description on the dataset page: https://huggingface.co/datasets/Faizaniqbal/british-library-book-images.anemia-survey-dataset
Anemia Detection — Multi-Modal Clinical SEWA Rural Dataset
Organisation: SEWA Rural — Society for Education, Welfare and Action (Rural), Jhagadia, Gujarat, India
Dataset: sewa-rural-care/anemia-survey-dataset
Contact: sewarural@ymail.com
Version: 1.0 — July 2026
Dataset Summary
This dataset supports research into non-invasive, smartphone-based anemia
screening applicable to low-resource and rural healthcare settings. It was
collected by SEWA Rural — a non-profit… See the full description on the dataset page: https://huggingface.co/datasets/sewa-rural-care/anemia-survey-dataset.danbooru2025-metadata
🎨 Danbooru 2025 Metadata
Latest Post ID: 9,158,800
(as of Apr 16, 2025)
📁 About the DatasetThis dataset provides structured metadata for user-submitted images on Danbooru, a large-scale imageboard focused on anime-style artwork.
Scraping began on January 2, 2025, and the data are stored in Parquet format for efficient programmatic access.Compared to earlier versions, this snapshot includes:
More consistent tag history tracking
Better coverage of older or previously… See the full description on the dataset page: https://huggingface.co/datasets/trojblue/danbooru2025-metadata.TASTE
TASTE: Human Preferences for Design-Quality Image Comparison
This dataset is the human-evaluation corpus released alongside the
TASTE preference model. It contains panel rankings of generated
images across multiple quality dimensions — both aesthetic (does the
image look good?) and description-faithfulness (does the image match
what the prompt describes?) — plus a per-image hallucination
judgement.
Quick stats
Table
Rows
Notes
prompts.parquet
~200
one… See the full description on the dataset page: https://huggingface.co/datasets/purvanshi/TASTE.PovertyMap
PovertyMap-Wilds: Poverty mapping across different countries
Description
This is a processed version of LandSat 5/7/8 satellite imagery originally from Google Earth Engine under the names LANDSAT/LC08/C01/T1_SR,LANDSAT/LE07/C01/T1_SR,LANDSAT/LT05/C01/T1_SR,
nighttime light imagery from the DMSP and VIIRS satellites (Google Earth Engine names NOAA/DMSP-OLS/CALIBRATED_LIGHTS_V4 and NOAA/VIIRS/DNB/MONTHLY_V1/VCMSLCFG)
and processed DHS survey metadata obtained from… See the full description on the dataset page: https://huggingface.co/datasets/1aurent/PovertyMap.relaion2b-natural-embeddings
LAION-Natural Embeddings: CLIP ViT-H/14 Features for ~500M Natural Photographs (CCN 2025, Roth & Hebart)
LAION-Natural Embeddings provides pre-computed CLIP ViT-H/14 embeddings for ~500 million natural photographs from ReLAION-2B, filtered using the LAION-Natural naturalness classifier (score > 0.7).
Also known as: LAION-Natural Embeddings · ReLAION-Natural Embeddings · LAION-2B-Natural Embeddings
Part of the LAION-Natural dataset family, introduced in: How to sample the… See the full description on the dataset page: https://huggingface.co/datasets/andropar/relaion2b-natural-embeddings.britannica-illustrated-pages
Britannica Illustrated Pages
115,293 illustrated pages from scanned volumes of the Encyclopaedia Britannica, 1st edition
(1768–71) to 14th (1929), selected by a page classifier from 975,345 pages in 1,160 volumes
(838 Internet Archive items). A second config carries the classifier
score, OCR word count and provenance for every one of the 975,345 pages.
Two things the scan showed:
82% of the illustrated pages are text pages (≥100 OCR words) — figures, diagrams and
engravings set… See the full description on the dataset page: https://huggingface.co/datasets/biglam/britannica-illustrated-pages.GTSRB
Dataset Card for GTSRB
Dataset Summary
The German Traffic Sign Benchmark is a multi-class, single-image classification challenge held at the International Joint Conference on Neural Networks (IJCNN) 2011. We cordially invite researchers from relevant fields to participate: The competition is designed to allow for participation without special domain knowledge. Our benchmark has the following properties:
Single-image, multi-class classification problem
More than 40… See the full description on the dataset page: https://huggingface.co/datasets/bazyl/GTSRB.vecforge-paper-corpus
Note (rebuild in progress): figures are being re-extracted with a fixed extractor (cleaner crops). The image-preview config (Parquet with an inline column + difficulty/type/score labels) returns after re-classification. The config (paper metadata + links) is live now.
VecForge Paper Corpus
A pristine, deduplicated collection of 59,732 top-venue AI/ML/CV/NLP/Robotics papers (2020-2024) with
every captioned figure and full paper text, for research on figure understanding… See the full description on the dataset page: https://huggingface.co/datasets/debajyotidasgupta/vecforge-paper-corpus.Icarus-dataset
Icarus
A unified multi-modal curriculum dataset for evolutionary neural architecture search. Every row is one self-contained Task = {meta, support, query}, where support and query are lists of (input_Field, output_Field) pairs. The inner loop trains on support; fitness is scored on query. Support is non-empty for every task. Encoders read the Field descriptor (axes, value_type, n_classes, value_range, mask); mask is True where a value is padding/ignored. meta.class_names, when… See the full description on the dataset page: https://huggingface.co/datasets/Ardea/Icarus-dataset.Comiman-Dataset
Comiman Dataset
Attribution is required for every use: Comiman Dataset by Waheed (huggingface.co/Waheed786dar) - https://huggingface.co/datasets/Waheed786dar/Comiman-Dataset
A license-gated comics and manga page corpus built for training a model that can plan and draw full
comic/manga series (the planned model: Waheed786dar/Comiman).
Every book passed an automatic license gate (Creative Commons / CC0 / Public Domain Mark metadata, or a
public-domain claim limited to works… See the full description on the dataset page: https://huggingface.co/datasets/Waheed786dar/Comiman-Dataset.gelsight-mini-pretrain
GelSight Mini Pretrain
~853K GelSight Mini tactile RGB frames, 12 public sources, one parquet schema. Built for self-supervised representation learning (VAE / MAE / SimCLR / DINO) — every frame contact-filtered, channel-normalized, and re-encoded as JPEG q92.
Frames
Sources
Real
536K
FoTA (labeled+unlabeled), 3DCal, FEATS, GelSLAM, TactileTracking, RTM, FeelAnyForce, UniT, TacQuad
Sim
317K
sim_tactile_mnist, sim_starstruck (Taxim-rendered, Mini-calibrated)
NC… See the full description on the dataset page: https://huggingface.co/datasets/yxma/gelsight-mini-pretrain.GeoMeld
🌍 GeoMeld Multi-Modal Earth Observation Dataset (WebDataset)
GeoMeld is a large-scale multi-modal remote sensing dataset introduced in our CVPRW 2026 paper on semantically grounded foundation modeling.
GeoMeld contains approximately 2.5 million spatially aligned samples spanning heterogeneous sensing modalities and spatial resolutions, paired with semantically grounded captions generated through an agentic pipeline.
The dataset is designed to support multimodal representation… See the full description on the dataset page: https://huggingface.co/datasets/vimageiitb/GeoMeld.liquidrandom-data
liquidrandom-data
Diverse seed data for ML/LLM training data generation pipelines.
Used by the liquidrandom Python package.
Dataset Summary
This dataset contains 520,080 seed data samples across 24 categories,
generated using a hierarchical taxonomy tree approach with LLM-based quality validation
and fuzzy deduplication. Data is stored as Parquet with zstd compression.
Categories
Category
Samples
File
Coding Tasks
30,069… See the full description on the dataset page: https://huggingface.co/datasets/mlech26l/liquidrandom-data.Flickr-Geo
Dataset Card for Flickr-Geo
217.646.487 images with lat lon coordinates from Flickr. This repo is a filtered version of https://huggingface.co/datasets/bigdata-pw/Flickr for all rows containing a valid lat lon pair.
Load and Visualize
Load the data (2s on my system) with DuckDB:
import duckdb
df = duckdb.sql("""
SELECT CAST(latitude AS DOUBLE) AS latitude,
CAST(longitude AS DOUBLE) AS longitude
FROM 'reduced_flickr_data/*.parquet'
""").df()
df… See the full description on the dataset page: https://huggingface.co/datasets/do-me/Flickr-Geo.coyo-labeled-300m
Dataset Card for COYO-Labeled-300M
Dataset Summary
COYO-Labeled-300M is a dataset of machine-labeled 300M images-multi-label pairs. We labeled subset of COYO-700M with a large model (efficientnetv2-xl) trained on imagenet-21k. We followed the same evaluation pipeline as in efficientnet-v2. The labels are top 50 most likely labels out of 21,841 classes from imagenet-21k. The label probabilies are provided rather than label so that the user can select threshold of their… See the full description on the dataset page: https://huggingface.co/datasets/kakaobrain/coyo-labeled-300m.openbrush-landscapes
OpenBrush Landscapes
Every landscape painting from OpenBrush-75K — across all artists, movements, and centuries. Largest single-genre subset.
Curated subset of jaddai/openbrush. Same CC0 license, same caption schema, same VLM (Qwen3-VL-30B-A3B). This subset exists so you don't have to download 75,313 images to get to the 12,612 you actually want.
Why this subset
Every landscape across the parent dataset's full range — Romantic wildernesses, Impressionist… See the full description on the dataset page: https://huggingface.co/datasets/jaddai/openbrush-landscapes.trade_vision_dataset
TradeVision: Hierarchical Physical Business & Multimodal Retail Provenance Dataset
This dataset is continuously seeded from OpenStreetMap, matched to Google Place IDs, harvested for temporal store photos, and enriched with zero-shot computer vision using Hugging Face Hub native pipelines.
Dataset Structure
The dataset is partitioned into three relational subsets loadable via Hugging Face datasets:
from datasets import load_dataset
# 1. Load Canonical Businesses… See the full description on the dataset page: https://huggingface.co/datasets/drksci/trade_vision_dataset.inline-digital-holography-v3
Dataset Card for Synthetic Inline Holographical Images v3 (224px Highly Diverse)
This dataset provides synthetic image triplets representing inline holographical imaging in a simulated environment. This version (v3) uses a native 224x224 resolution optimized for modern Vision Transformers (ViT, Swin) and contains 25,000 samples across 8 noise configurations.
Each data sample consists of:
An object-domain field (ground truth),
Its corresponding forward-propagated hologram (the… See the full description on the dataset page: https://huggingface.co/datasets/gokhankocmarli/inline-digital-holography-v3.Bodhisetu
Bodhisetu
A multimodal cultural-heritage dataset from India, collected via the SMRITI field-data-collection platform.
Overview
Bodhisetu documents 3,000+ heritage entities across 6 Indian languages (Tamil, Kannada, Assamese, Maithili, Bodo, Nepali) and 14 taluks spanning 5 states. Each entity is captured as one or more images and short-form videos, accompanied by field descriptions.
This release contains 5,753 resources (4,461 images + 1,292 videos, ~91 GB)… See the full description on the dataset page: https://huggingface.co/datasets/deb0naire/Bodhisetu.meowcat-predictions
MeowCat cell-type predictions on TCGA-LUAD and CPTAC-CCRCC
Per-pixel cell-type predictions generated by MeowCat on
H&E whole-slide images from two public cohorts:
Cohort
Tissue
Samples
h5ad payload
TCGA-LUAD
Lung adenocarcinoma
531
~60 GB
CPTAC-CCRCC
Clear-cell renal cell carcinoma
831
~93 GB
File layout
composition.parquet # long format: sample × cell_type → count, fraction
metadata.parquet # sample_id, cohort, patient_id, n_pixels… See the full description on the dataset page: https://huggingface.co/datasets/liranmao/meowcat-predictions.trace-rx-eval-predictions
TRACE-RX Evaluation Predictions
Per-image detector scores from an independent evaluation of the two TechJam 2026 TRACE-RX
detectors, run 30 Aug – 1 Sep 2026.
No images here. Every file contains scores, labels, asset ids and transform names only — this is
derived evaluation metadata, not a redistribution of any source imagery. The underlying corpora
(Joshyxwa/data_draft, Joshyxwa/techjam2026, techjam-aigc/wildfake-eval-subset) keep their own
terms, and data_draft's WildFake rows… See the full description on the dataset page: https://huggingface.co/datasets/joelleoqiyi/trace-rx-eval-predictions.house_kg_full_dataset
house.kg — Kyrgyzstan Real Estate (multimodal)
A complete snapshot of house.kg, the largest real-estate
board in Kyrgyzstan: every sale and rental listing, with coordinates, prices, seller
identities, agency ratings, reviews — and 227,294 photographs.
Field names are English; values are kept in the original language (Russian/Kyrgyz),
exactly as the site renders them.
💻 Scraper source code on GitHub →
The complete, open scraper that produced this dataset —… See the full description on the dataset page: https://huggingface.co/datasets/aiacademy-kg/house_kg_full_dataset.ODELIA-Challenge-2025
ODELIA Challenge Dataset
This dataset is part of the ODELIA project, a European Horizon initiative focused on developing privacy-preserving, AI-driven diagnostic tools using swarm learning.
The dataset provided here represents a curated subset of data from the broader ODELIA consortium. It is designed to facilitate the development, benchmarking, and validation of AI algorithms that can operate effectively across a range of heterogeneous clinical settings.
The dataset contains… See the full description on the dataset page: https://huggingface.co/datasets/ODELIA-AI/ODELIA-Challenge-2025.openbrush-75k
OpenBrush-75K
A curated dataset of 75,313 public domain artworks with rich, structured VLM-generated captions designed for training image generation models, fine-tuning vision-language models, and art analysis research.
Dataset Description
OpenBrush-75K contains high-quality reproductions of paintings from the Western art canon, spanning from the Renaissance to the early 20th century. Each image is paired with a detailed structured caption generated by a vision-language… See the full description on the dataset page: https://huggingface.co/datasets/Trever896/openbrush-75k.nesteo-prototype
NestEO: Modular and Hierarchical EO Dataset Framework
NestEO is a hierarchical, resolution-aligned, UTM-based nested grid dataset framework supporting general-purpose, multi-scale multimodal Earth Observation workflows. Built from diverse EO sources and enriched with metadata for landcover, climate zones, and population, it enables scalable, representative and progressive sampling for AI4EO.
Grid Levels: 120000m, 12000m, 2400m, 1200m, 600m, 300m, 150mGrid Metadata: ESA WorldCover… See the full description on the dataset page: https://huggingface.co/datasets/nesteo-datasets/nesteo-prototype.