datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
food101
Dataset Card for Food-101
Dataset Summary
This dataset consists of 101 food categories, with 101'000 images. For each class, 250 manually reviewed test images are provided as well as 750 training images. On purpose, the training images were not cleaned, and thus still contain some amount of noise. This comes mostly in the form of intense colors and sometimes wrong labels. All images were rescaled to have a maximum side length of 512 pixels.
Supported Tasks and… See the full description on the dataset page: https://huggingface.co/datasets/ethz/food101.imagenet_1k_resized_256
Dataset Card for "imagenet_1k_resized_256"
Dataset summary
The same ImageNet dataset but all the smaller side resized to 256.
A lot of pretraining workflows contain resizing images to 256 and random cropping to 224x224, this is why 256 is chosen.
The resized dataset can also be downloaded much faster and consume less space than the original one.
See here for detailed readme.
Dataset Structure
Below is the example of one row of data. Note that the labels in… See the full description on the dataset page: https://huggingface.co/datasets/evanarlian/imagenet_1k_resized_256.eurosat
Dataset Card for EuroSAT
Dataset Source
Paper with code
Usage
from datasets import load_dataset
dataset = load_dataset('tranganke/eurosat')
Data Fields
The dataset contains the following fields:
image: An image in RGB format.
label: The label for the image, which is one of 10 classes:
0: annual crop land
1: forest
2: brushland or shrubland
3: highway or road
4: industrial buildings or commercial buildings
5: pasture land
6: permanent crop land… See the full description on the dataset page: https://huggingface.co/datasets/tanganke/eurosat.Language-Grounded_Sparse_Encoder_Training
Language-Grounded Sparse Encoder (LanSE) — Training Data
This repository hosts the AI-generated images and human annotation datasets accompanying the paper:
Human-like Content Analysis for Generative AI with Language-Grounded Sparse Encoders
Yiming Tang, Arash Lagzian, Srinivas Anumasa, Qiran Zou, Yingtao Zhu, Ye Zhang, Trang Nguyen, Yih-Chung Tham, Ehsan Adeli, Ching-Yu Cheng, Yilun Du, Dianbo Liu
National University of Singapore · Tsinghua University · Stanford University ·… See the full description on the dataset page: https://huggingface.co/datasets/DesmondYMTang2024/Language-Grounded_Sparse_Encoder_Training.eurosatRedistributed without modification from https://github.com/phelber/EuroSAT.
EuroSAT100 is a subset of EuroSATallBands containing only 100 images. It is intended for tutorials and demonstrations, not for benchmarking.
zju-eye-pretrain
ZJU Eye-Pretrain Dataset
Unified multi-source ophthalmological imaging dataset for foundation model pretraining and downstream tasks.
1.18M images spanning 57 cohorts with a strict 41-column unified manifest schema.
Composition
1,179,458 images · 57 cohorts · 8 modalities, under one strict 41-column unified manifest schema. Stored as HF parquet shards in data/{private_topcon, public_fundus, public_oct, public_new}/; loadable by batch or by single cohort (57… See the full description on the dataset page: https://huggingface.co/datasets/MaybeRichard/zju-eye-pretrain.olmoearth-paper-embeddings
OlmoEarth — Foundation-Model Embeddings for Paper Table 2
This dataset contains pre-extracted embeddings from 26 Earth-observation
foundation models evaluated on the 24 downstream tasks that make up
Table 2 of the OlmoEarth paper:
OlmoEarth: Stable Latent Image Modeling for Multimodal Earth Observation
AI2, 2025. arXiv:2511.13655.
For every supported (model, task) pair we ran the model's encoder over the
task's train / validation / test splits with the paper-best… See the full description on the dataset page: https://huggingface.co/datasets/allenai/olmoearth-paper-embeddings.Kino-Fail
Kino-Fail
Kino-Fail is a counterfactual benchmark for recovery-relevant failure attribution in quadrupedal navigation. The ICRA paper-facing release contains 11 physics and sensing interventions, 191 RTX/PBR scenes, 11,418 matched groups, and 22,836 accepted physical units.
Release status: complete.
Data organization
data/scale/<scene>/: five-frame front-camera RGB and 21x19 proprioception windows.
data/conflict/<scene>/: T2 visual-decisive and T3… See the full description on the dataset page: https://huggingface.co/datasets/EurekaZang123/Kino-Fail.imagenet-1k-vl-enriched
Visualize on Visual Layer
Imagenet-1K-VL-Enriched
An enriched version of the ImageNet-1K Dataset with image caption, bounding boxes, and label issues!
With this additional information, the ImageNet-1K dataset can be extended to various tasks such as image retrieval or visual question answering.
The label issues helps to curate a cleaner and leaner dataset.
Description
The dataset consists of 6 columns:
image_id: The original filename of the image from… See the full description on the dataset page: https://huggingface.co/datasets/visual-layer/imagenet-1k-vl-enriched.echoxflow
EchoXFlow - 2-D B-mode Ultrasound Dataset
This dataset is a zea-format (HDF5) conversion of the EchoXFlow
2d_brightness_mode recordings, hosted at
zeahub/echoxflow.
Conversion
This dataset was converted to zea format and uploaded using the
zea data converter:
python -m zea.data.convert echoxflow <src> <dst>
Dataset structure
<exam_id>/
<recording_id>.hdf5
...
Each HDF5 file follows the zea data format.
EuroSAT_RGB
EuroSAT RGB
EUROSAT RGB is the RGB version of the EUROSAT dataset based on Sentinel-2 satellite images covering 13 spectral bands and consisting of 10 classes with 27000 labeled and geo-referenced samples.
Paper: https://arxiv.org/abs/1709.00029
Homepage: https://github.com/phelber/EuroSAT
Description
The EuroSAT dataset is a comprehensive land cover classification dataset that focuses on images taken by the ESA Sentinel-2 satellite. It contains a total of 27… See the full description on the dataset page: https://huggingface.co/datasets/blanchon/EuroSAT_RGB.Egocentric_10K_Evaluation
Dataset Card for Egocentric_10K_Evaluation
This is a FiftyOne dataset with 30000 samples.
Installation
If you haven't already, install FiftyOne:
pip install -U fiftyone
Usage
import fiftyone as fo
from fiftyone.utils.huggingface import load_from_hub
# Load the dataset
# Note: other available arguments include 'max_samples', etc
dataset = load_from_hub("Voxel51/Egocentric_10K_Evaluation")
# Launch the App
session = fo.launch_app(dataset)
Dataset… See the full description on the dataset page: https://huggingface.co/datasets/Voxel51/Egocentric_10K_Evaluation.laion2b-en-a65_cogvlm2-4bit_captions
Abstract
This dataset contains image captions for the laion2B-en aesthetics>=6.5 image dataset using CogVLM2-4bit with the "laion-pop"-prompt to generate captions which were "likely" used in Stable Diffusion 3 training. From these image captions new synthetic images were generated using stable-diffusion-3-medium (batch-size=8).
The synthetic images are best viewed locally by cloning this repo with:
git lfs install
git clone… See the full description on the dataset page: https://huggingface.co/datasets/GeroldMeisinger/laion2b-en-a65_cogvlm2-4bit_captions.EEG_Image_decode
EEG Image Decode — Dataset and Checkpoints
This dataset accompanies the NeurIPS 2024 paper:
Visual Decoding and Reconstruction via EEG Embeddings with Guided Diffusion Dongyang Li · Chen Wei · Shiying Li · Jiachen Zou · Quanying Liu
It packages the preprocessed EEG recordings, stimulus-image visual features, VAE latent codes, trained EEG embeddings, fine-tuned checkpoints, and generated images needed to reproduce both the image retrieval and image reconstruction experiments.… See the full description on the dataset page: https://huggingface.co/datasets/LidongYang/EEG_Image_decode.eurosat-rgb
EuroSat (RGB)
Description
A dataset based on Sentinel-2 satellite images covering 13 spectral bands and consisting of 10 classes with 27000 labeled and geo-referenced samples. This is the RGB version of the dataset with visible bands encoded as JPEG images.
The dataset does not have any default splits. Train, validation, and test splits were based on these definitions here… See the full description on the dataset page: https://huggingface.co/datasets/timm/eurosat-rgb.emnist-letters-tiny
Dataset Card for EMNIST-Letters-10k
A random subset of the train and test splits from the letters portion of EMNIST
This is a FiftyOne dataset with 10000 samples.
Installation
If you haven't already, install FiftyOne:
pip install -U fiftyone
Usage
import fiftyone as fo
from fiftyone.utils.huggingface import load_from_hub
# Load the dataset
# Note: other available arguments include 'max_samples', etc
dataset =… See the full description on the dataset page: https://huggingface.co/datasets/Voxel51/emnist-letters-tiny.mixlora-eval-data
🚀 MixLoRA Evaluation Data
This dataset is the held-out multimodal evaluation suite used in
Multimodal Instruction Tuning with Conditional Mixture of LoRA (ACL 2024).
It bundles 9 instruction-formatted tasks (mm_tasks/) plus the MME benchmark
(mme/) used to evaluate MixLoRA and baseline models in the paper.
The 9 tasks in mm_tasks/ are the zero-shot / held-out task split from
Vision-Flan. MME is a
separate benchmark, evaluated independently.
Dataset Structure… See the full description on the dataset page: https://huggingface.co/datasets/yingss/mixlora-eval-data.Danbooru-WD-EVA-EmbeddingsThis dataset includes WD EVA v2 large embeddings for danbooru images. Pixiv will be added later. Tensors under w are direct outputs and match indexes for WD EVA model. Tensors under e are from WD EVA as well however these strips the classifiaction head, they are smaller and suitable for deduplication computing for example.
The indexes of w, e and f (filename) match.
WD EVA Model: https://huggingface.co/SmilingWolf/wd-eva02-large-tagger-v3
efficientnet-v2-l-adv-dataset
Perturb Adversarial Images
Verified adversarial examples for efficientnet_v2_l (torchvision/EfficientNet_V2_L_Weights.IMAGENET1K_V1), produced by the
Perturb network. Each row is one clean image together with all of its
verified adversarial versions: images that are imperceptibly different from the original
(L∞ ≤ 0.03 in [0,1] pixel scale) yet change the model's top-1 prediction.
This dataset grows continuously. New rows are appended as the network produces them and uploaded in… See the full description on the dataset page: https://huggingface.co/datasets/perturb-ai/efficientnet-v2-l-adv-dataset.lgg-mri-segmentation-research
LGG Brain MRI Segmentation with Genomic Clusters
This repository provides a Patient-Centric version of the Lower-Grade Glioma (LGG) Segmentation dataset. While other versions of this data exist, they often treat slices as independent images. This version preserves the 3D patient volume and integrates all genomic/clinical labels directly into a multimodal-ready format.
🌟 Why This Version?
Developed for Multimodal AI Research, this dataset addresses several limitations… See the full description on the dataset page: https://huggingface.co/datasets/Ehsan-rmz/lgg-mri-segmentation-research.ops-eval
OPS-Eval: Leakage-Resistant Evaluation for Optical Pooled Screens
Benchmark artifacts for evaluating representation learning on pooled CRISPR
microscopy data. This dataset accompanies a submission to the NeurIPS 2026
Evaluations and Datasets Track.
Contents
Directory/File
Description
Size
montages/
Per-gene montage images (4 channels x 2 phases, ~10 PNGs per gene)
~66 GB
cell_embeddings/
Pre-extracted 512-dim cell embeddings per sgRNA (.npz)
~15 GB… See the full description on the dataset page: https://huggingface.co/datasets/cspeters119/ops-eval.SwissCrop25
SwissCrop25
A national benchmark dataset for operational crop mapping in Switzerland, providing Sentinel-2
time series, daily temperature data, and parcel-level crop type labels across seven growing
seasons (2019–2025).
Introduced in: SwissCrop25: A National Multi-Year Benchmark for Operational Crop Mapping
(TerraBytes II Workshop, ECCV 2026) — [Paper] [Code] [Team]
Highlights
Nationwide coverage of Switzerland (41,285 km²)
Seven growing seasons (2019–2025)
73… See the full description on the dataset page: https://huggingface.co/datasets/EOA-team/SwissCrop25.e621_newest
E621 Dataset Newest Supplement
This is the newest supplement dataset of e621.net. And only the newest data are up-to-date-ly maintained here, to make sure you can get all the newest data from huggingface instead of e621 site.
If you are looking for some old data, just see: boxingscorpionbagel/e621-2024
All the file types we kept in this dataset: gif, jpg, mp4, png, swf, webm
How to Painlessly Use This
Use cheesechaser to quickly get images from this repository and… See the full description on the dataset page: https://huggingface.co/datasets/deepghs/e621_newest.EmoVerse
EmoVerse
EmoVerse is a visual emotion dataset for affective image understanding. The dataset is organized around eight emotion categories:
Amusement, Anger, Awe, Contentment, Disgust, Excitement, Fear, and Sadness.
The released package contains annotation files and Parquet shards for the image records and annotations. The Parquet rows store file-level data: each row describes one packed file and includes both metadata and the file content as a binary column.… See the full description on the dataset page: https://huggingface.co/datasets/alkalol/EmoVerse.Wake-Vision
Dataset Card for Wake Vision
Dataset Description
"Wake Vision" is a large, high-quality dataset featuring over 6 million images, significantly exceeding the scale and diversity of
current tinyML datasets (100x). This dataset includes images with annotations of whether each image contains a person. Additionally,
it incorporates a comprehensive fine-grained benchmark to assess fairness and robustness, covering perceived gender, perceived age,
subject distance, lighting… See the full description on the dataset page: https://huggingface.co/datasets/Harvard-Edge/Wake-Vision.StreetView360AtoZStreetView 360X is a dataset containing 6342 360 degree equirectangular street view images randomly sampled and downloaded from Google Street View. It is published as part of the paper "StreetView360X: A Location-Conditioned Latent Diffusion Model for Generating Equirectangular 360 Degree Street Views" (Princeton COS Senior Independent Work by Everett Shen). Images are labelled with their capture coordinates and panorama IDs. Scripts for extending the dataset (i.e. fetching additional images)… See the full description on the dataset page: https://huggingface.co/datasets/everettshen/StreetView360AtoZ.Holistic-Processing-Illusion-Faces
HoloFaceIllusion-Bench-EEG
A large-scale benchmark of holistic-face illusion stimuli for testing
human-vs-DNN alignment on configural face processing and for paired
EEG-decoder evaluation. Built entirely with classical CV
(dlib landmarks + MediaPipe Face Mesh + InsightFace gender/age +
OpenCV Poisson cloning + Reinhard LAB colour transfer) — no neural
networks or generative AI are used to create any pixel.
Three paradigms are included:
Paradigm
Cases
Conditions per case… See the full description on the dataset page: https://huggingface.co/datasets/Enhui-1/Holistic-Processing-Illusion-Faces.si_us_revolutionary_era_collections
Dataset Card for Smithsonian American Revolutionary Era Collections
Dataset Summary
A specially selected subset of the Smithsonian’s Open Access collections covering objects from 1770–1810 selected for the Revolution Crossroads project in honor of the 250th anniversary of the founding of the United States. Drawn from four museums—the National Museum of American History, National Postal Museum, Smithsonian American Art Museum, and National Portrait Gallery—the… See the full description on the dataset page: https://huggingface.co/datasets/RevolutionCrossroads/si_us_revolutionary_era_collections.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.synthetic_us_passports_easy
Dataset Card for synthetic_us_passports
This is a FiftyOne dataset with 9750 samples.
Installation
If you haven't already, install FiftyOne:
pip install -U fiftyone
Usage
import fiftyone as fo
from fiftyone.utils.huggingface import load_from_hub
# Load the dataset
# Note: other available arguments include 'max_samples', etc
dataset = load_from_hub("Voxel51/synthetic_us_passports_easy")
# Launch the App
session = fo.launch_app(dataset)
Based on the… See the full description on the dataset page: https://huggingface.co/datasets/Voxel51/synthetic_us_passports_easy.