Skip to content

Datasets

Hugging Face datasets with plain-language descriptions: what's inside and what models to train on them. 4,788 in the catalog so far, with new ones added every hour.

Data type

For which models

Find a dataset for your goal

Answered by Gemini 4 · the chat is kept only in this tab

DatasetHuggingFaceVLA/libero

Panda robot data for LIBERO tasks

Panda robot manipulation episodes: 1,693 episodes, 273k frames, and 40 tasks for training robotics policies.

  • Robotics
14K 72100K–1M rows
DatasetShareGPT4Video/ShareGPT4Video

ShareGPT4Video: video captions

4.8M multimodal GPT-4-Vision video descriptions for LVLM and text-to-video models.

  • Visual question answering
  • Question answering
  • English
14K 21610K–100K rows
Datasetrafaelpadilla/coco2017

COCO 2017 Object Detection Mirror

A complete set of COCO 2017 images and annotations for training and evaluating computer vision models.

  • Object detection
  • English
13.6K 32100K–1M rows
Dataseteltorio/ROCOv2-radiology

79.8k radiology images with captions

Multimodal set of radiology scans with captions and medical concepts from PMC.

  • English
13.6K 9110K–100K rows
DatasetShadenA/MathNet

Multilingual math problems with images

Problems from countries and competitions with images and text for QA and solution generation.

  • Question answering
  • Text generation
  • Image → text
  • English
  • Portuguese
  • Spanish
  • +9
13.3K 10510K–100K rows
DatasetSandflow/sand3d-automate

Parametric 3D objects and mechanisms as code

3,169 3D models with Python code in sand3d, OpenUSD export, and renders for text-to-3D and image-to-3D.

  • Text → 3d
  • Image → 3d
13.2K 61K–10K rows
DatasetOpenEnvisionLab/WorldAtlasTop pick

WorldAtlas benchmark for evaluating generative world models

A large collection of labeled images and reference videos in English to check consistency and physical plausibility of scene generators.

  • Other
  • English
13K 810K–100K rows🔒 on request
DatasetJailbreakV-28K/JailBreakV-28k

JailBreakV-28K: MLLM jailbreak benchmark

A collection of images and texts for testing multimodal LLM robustness against jailbreak attacks.

  • Text generation
  • Question answering
12.9K 9210K–100K rows
Datasetwikimedia/wit_base

Multilingual image-text pairs from Wikipedia

Image and text connections from Wikipedia across many languages for training and evaluating multimodal retrieval.

  • Image → text
  • Text retrieval
  • Afrikaans
  • Aragonese
  • Arabic
  • +9
12.5K 77100K–1M rows
Datasetanthonytec2/OctoSenseTop pick

OctoSense multi-sensor driving dataset

8.43 TB of synchronized sensor data for autonomous driving and robotics.

  • Depth estimation
  • Image segmentation
  • Robotics
12.4K 38up to 1K rows
Datasetosunlp/Multimodal-Mind2Web

Multimodal dataset for Mind2Web web agents

HTML pages with screenshots and instructions for training web agents to perform tasks on websites.

  • English
12.4K 10210K–100K rows
Datasetjiawennnn/STimage-1K4M

Histopathology and gene expression for transcriptomics

High-resolution tissue images and gene expression data for biomedical models.

  • Image feature extraction
  • Image segmentation
  • Image classification
  • English
12.4K 36100B–1T rows
Datasetnyu-visionx/CV-Bench

CV-Bench visual benchmark for multimodal LLMs

1K to 10K test examples with images and questions to evaluate vision capabilities in multimodal models.

  • Visual question answering
  • English
12.2K 491K–10K rows
DatasetUrbanSyn/UrbanSyn

Photorealistic synthetic road scenes with annotations

Over 7,500 synthetic road scene images with depth, segmentation, and object annotations for autonomous driving.

  • Object detection
  • Image segmentation
  • Depth estimation
  • English
11.9K 3110K–100K rows
DatasetTsinghuaC3I/MedXpertQA

MedXpertQA expert medical questions

4,460 challenging medical questions spanning text and multimodal tasks with clinical images.

  • Table question answering
  • Question answering
  • Text generation
  • English
11.8K 681K–10K rows
DatasetRapidata/text-to-image-human-feedback-benchmarkTop pick

Text-to-image human preference dataset

Largest open human preference dataset for text-to-image models with 7.7M votes.

  • Text → image
  • English
11.5K 26100K–1M rows
Datasetshowlab/OmniConsistency

Paired images across 22 art styles

Dataset of source-to-stylized image pairs with text prompts for training style transfer.

    10.9K 391K–10K rows