datasets
Training and evaluation data, with the modality, task and licence stated up front. Listed live from the Hugging Face Hub.
witmini-imagenet
Dataset Description
A mini version of ImageNet-1k with 100 of 1000 classes present.
Unlike some 'mini' variants this one includes the original images at their original sizes. Many such subsets downsample to 84x84 or other smaller resolutions.
Data Splits
Train
50000 samples from ImageNet-1k train split
Validation
10000 samples from ImageNet-1k train split
Test
5000 samples from ImageNet-1k validation split (all 50 samples per class)… See the full description on the dataset page: https://huggingface.co/datasets/timm/mini-imagenet.scannet_mini_val_set_suitemini-VTAB
Mini-VTAB
A collection of VTAB (Visual Task Adaptation Benchmark) datasets. We sampled 1K training samples and 1K testing samples for each task.
Tasks
datasets = [
"caltech101",
"cifar10",
"cifar100",
"dtd",
"flowers",
"pets",
"sun397",
"svhn",
"pcam",
"eurosat",
"resisc45",
"diabetic_retinopathy",
"clevr_count_all",
"clevr_closest_object_distance",
"dmlab",
"dsprites_label_x_position"… See the full description on the dataset page: https://huggingface.co/datasets/antofuller/mini-VTAB.reachy-mini-wall-data
Reachy Mini — wall data (public)
posts.json for the Reachy Mini community wall: the AI-filtered posts shown publicly,
aggregated from Bluesky, YouTube, LinkedIn, TikTok, X and Reddit by the social-wall pipeline.
Fetch it directly (CORS-enabled) from any static site:
const url = "https://huggingface.co/datasets/pollen-robotics/reachy-mini-wall-data/resolve/main/posts.json";
const posts = await (await fetch(url)).json();
Each item: id, platform, author, handle, avatar, text… See the full description on the dataset page: https://huggingface.co/datasets/pollen-robotics/reachy-mini-wall-data.minimax-h3-soup
MiniMax H3 Soup
Reproducibility archive for a local ComfyUI MiniMax H3 Ref2V benchmark on an RTX 3090.
What is included
Original benchmark workflow graph (source_prompt.json), manifest, and result table.
Every one-second MP4 from the original C1-C11 benchmark grid and its Euler
repeat sweep. The separate
duration experiments are intentionally not included.
Labeled C1-C11 visual contact sheets, Ref2VA stock-control sheets, and a
static render-time summary chart.… See the full description on the dataset page: https://huggingface.co/datasets/badincite/minimax-h3-soup.open-imagesVisualProbe_trainmini-VTAB-corruptions
mini-VTAB-C
A collection of VTAB (Visual Task Adaptation Benchmark) datasets. We sampled 1K training samples and 1K testing samples for each task. For each test set, we apply all 15 corruption types from ImageNet-C.
Tasks
datasets = [
"caltech101",
"cifar10",
"cifar100",
"dtd",
"flowers",
"pets",
"sun397",
"svhn",
"pcam",
"eurosat",
"resisc45",
"diabetic_retinopathy",
"clevr_count_all"… See the full description on the dataset page: https://huggingface.co/datasets/antofuller/mini-VTAB-corruptions.conus
Contiguous UNited States AOI
Randomly sampled 5000 data tiles (3000 for training, and 1000 for validation and testing).
Occasionally, not all tiles were available for all datasets. We include a detailed breakdown of the number of chips per dataset below.
Datakind
Chips
s1grd-2020
5000
gssic
5000
gunw_2020-04-01_2020-06-30
4138
gunw_2020-07-01_2020-09-30
4171
gunw_2020-08-01_2020-10-31
4166
gunw_2020-01-01_2020-03-31
4097
s2rgbm-2020
5000
biomass-2020
5000… See the full description on the dataset page: https://huggingface.co/datasets/M3LEO-miniset/conus.middleeast
Middle East AOI
Randomly sampled 5000 data tiles (3000 for training, and 1000 for validation and testing).
Occasionally, not all tiles were available for all datasets. We include a detailed breakdown of the number of chips per dataset below.
Datakind
Chips
s1grd-2020
5000
gssic
4829
gunw_2020-04-01_2020-06-30
4641
gunw_2020-01-01_2020-03-31
4561
s2rgbm-2020
5000
biomass-2020
5000
esaworldcover-2020
5000
modis44b006veg
5000
ghsbuilts-2020
5000
srtmdem… See the full description on the dataset page: https://huggingface.co/datasets/M3LEO-miniset/middleeast.mini_imagenet
Dataset Card for "mini_imagenet"
More Information needed
minimind-v_dataset
Ⅰ 数据集
本轮训练用到的图文数据全部来自 ALLaVA-4V 系列。
相比以往从几份 LLaVA 衍生集拼接得到的数据，ALLaVA-4V 的质量更整齐、中英双语原生对照，细粒度描述也更充分。
它由两个子源构成：一份是 LAION 里挑出来的高质量图片（自然图像为主），一份是 VFLAN 指令流里挑出来的图片（文档、图表、合成场景居多）。
Pretrain（pretrain_i2t.parquet，约 127 万条 / ~64 万张唯一图像）
ALLaVA-Caption-LAION-4V 英/中：~47万 + ~44万ALLaVA-Caption-VFLAN-4V 英/中：~19万 + ~17万
任务形式为"请描述这张图片"类的单轮长描述，用于让模型建立视觉 token 到语言 token 的基础对齐。
SFT（sft_i2t.parquet，约 290 万条 / ~65 万张唯一图像）
ALLaVA-Instruct-LAION-4V 英/中：~47万 + ~47万… See the full description on the dataset page: https://huggingface.co/datasets/jingyaogong/minimind-v_dataset.mini-reachy-animation
Reachy Mini Animation Dataset
Multi-view renders of 85 emotional animations performed by the Reachy Mini
robot, paired with the full robot joint state for every single frame.
Source of the animations. The emotional animations rendered here come from the
official pollen-robotics/reachy-mini-emotions-library
dataset by Pollen Robotics. This dataset re-renders those emotions from 12 camera
angles (with 3 background variants) and pairs every frame with the robot's joint state.… See the full description on the dataset page: https://huggingface.co/datasets/BastienATOS/mini-reachy-animation.MiniCPM5-1B-atlas
juiceb0xc0de/MiniCPM5-1B-atlas
A brain atlas for openbmb/MiniCPM5-1B, a 1B on-device model with a 130k bilingual vocabulary. This is not a chat dataset or a benchmark. It is an internal-mechanics map, built by running activations through a corpus of prompts and scoring what each layer, component, head, and feature direction is doing.
If you want to know which parts of this model are safe to edit, where its output-vocabulary directions live, or which layers are carrying the most… See the full description on the dataset page: https://huggingface.co/datasets/juiceb0xc0de/MiniCPM5-1B-atlas.minimax_h3_avatar_500
Watch the full 500-video showcase on YouTube
MiniMax H3 Avatar 500
An image-to-video dataset pairing reference avatar images with detailed generation prompts and generated avatar videos. This release contains 500 curated examples in both a browsable raw layout and a typed Hugging Face dataset.
Version 1.0 · Released August 14, 2026
Dataset contents
Each example contains:
A 1024 × 1024 reference avatar image
A detailed English generation prompt
A generated 640 ×… See the full description on the dataset page: https://huggingface.co/datasets/oakmindai/minimax_h3_avatar_500.Radiology_mini0.33% sampled from https://huggingface.co/datasets/eltorio/ROCOv2-radiology
vqgan-pairs
VQGAN Pairs
This dataset contains ~2.4 million image pairs intended for improvement of image quality in VQGAN predictions. Each pair consists of:
A 512x512 crop of an image taken from Open Images.
A 256x256 image encoded and decoded using VQGAN, corresponding to the same image crop as the original.
This is the VQGAN implementation that was used for encoding and decoding: https://github.com/patil-suraj/vqgan-jax
License
This dataset is created using Open Images… See the full description on the dataset page: https://huggingface.co/datasets/dalle-mini/vqgan-pairs.OxHyperMinerals_MINIllava-instruct-mix-vsft-miniOriginally from https://huggingface.co/datasets/HuggingFaceH4/llava-instruct-mix-vsft but 0.33% randomnly sampled
VisualProbe_EasyVisualProbe_Hardcoco-val2017-mini500
COCO val2017 mini-500
A frozen, reproducible 500-image subset of COCO val2017 for fast object-detection
latency/throughput benchmarking on edge devices (e.g. LibreYOLO on NVIDIA Jetson), where the
full 5000-image val set is impractical.
Selection (deterministic)
The 500 images are sorted(COCO.getImgIds())[:500] of the official instances_val2017.json
(the 500 lowest image IDs) — identical to running the
vision-analysis-benchmark harness with --limit 500.… See the full description on the dataset page: https://huggingface.co/datasets/LibreYOLO/coco-val2017-mini500.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.VisualProbe_MediumVQA-RADmini-imagenetchina
China AOI
Randomly sampled 5000 data tiles (3000 for training, and 1000 for validation and testing).
Occasionally, not all tiles were available for all datasets. We include a detailed breakdown of the number of chips per dataset below.
Datakind
Chips
s1grd-2020
5000
gssic
5000
gunw_2020-01-01_2020-03-31
4354
s2rgbm-2020
5000
biomass-2020
5000
esaworldcover-2020
5000
modis44b006veg
5000
ghsbuilts-2020
5000
srtmdem
5000
We provide '.csv' files with… See the full description on the dataset page: https://huggingface.co/datasets/M3LEO-miniset/china.arbiter-mini
Arbiter-mini
A small, purpose-built image dataset of household items captured under controlled Raspberry Pi camera conditions and labeled for binary waste/recycle classification according to San Diego, CA municipal recycling rules. Built as deployment-condition training data for the Arbiter sorting system, intended to be used alongside TrashNet to close the domain gap between studio imagery and real Pi-camera inference.
Motivation
Models trained purely on TrashNet… See the full description on the dataset page: https://huggingface.co/datasets/aaryavlal/arbiter-mini.OmniEdit-miniMini version of TIGER-Lab/OmniEdit-Filtered-1.2M for rapid experimentation.
Script used:
from huggingface_hub import dataset_info, snapshot_download
import glob
from datasets import Dataset
import random
random.seed(2025)
def download_mini_omniedit_files():
repo_id = "TIGER-Lab/OmniEdit-Filtered-1.2M"
files = dataset_info(repo_id)
files = {f.rfilename for f in files.siblings if "data/" in f.rfilename}
files = sorted(list(files))
print(files[:5])
random.shuffle(files)… See the full description on the dataset page: https://huggingface.co/datasets/sayakpaul/OmniEdit-mini.