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

Datasetkakaobrain/coyo-700m

747M Image-Text Pairs

A large corpus of alt-text and image pairs from HTML for training multimodal models.

  • Text → image
  • Image → text
  • Zero shot classification
  • English
6.5K 177100M–1B rows
DatasetDelores-Lin/MDPBenchTop pick

Multilingual document parsing benchmark

The first benchmark for evaluating recognition and parsing of real digital and photographed documents across 6 languages.

  • Image → text
  • Chinese
  • English
  • Arabic
  • +9
6.3K 60
Datasetflaviagiammarino/vqa-rad

Radiology image question answering dataset

Question-answer pairs for MedPix medical images for VQA training and testing.

  • Visual question answering
  • English
5.9K 1051K–10K rows
DatasetUkrainianCatholicUniversity/rukopysTop pick

Annotated handwritten Ukrainian texts

The first major open dataset for Ukrainian handwritten text recognition spanning over a century.

  • Object detection
  • Image → text
  • Ukrainian
5.8K 2810K–100K rows
DatasetOmniGen2/X2I2

Multimodal Image Generation and Editing Dataset

Millions of English and Chinese examples for training image editing and generation models.

  • Image → image
  • Text → image
  • Multimodal
  • English
  • Chinese
5.7K 411M–10M rows
Datasetkurumikz/AnimeOmni

Anime characters with multi-level annotation and descriptions

Over 20 thousand safe-rated anime illustrations with detailed tags, emotions, and descriptions in three languages.

  • Text → image
  • Image → text
  • Image classification
  • English
  • Chinese
  • Japanese
5.5K 1310K–100K rows
Datasetyifanzhang114/MME-RealWorldTop pick

Multimodal Benchmark for Evaluating VLM Models

A challenging test set for checking visual understanding and visual question answering.

  • Multiple choice
  • Question answering
  • Visual question answering
  • English
5.3K 27100B–1T rows
Datasetvidore/colpali_train_set

ColPali training set for document retrieval

127K query-image pairs for training visual document retrieval.

  • Document question answering
  • Visual document retrieval
5.3K 93100K–1M rows
DatasetTHUIR/Qilin

Multimodal user session dataset for RAG and search

Large Chinese dataset (1M to 10M) for search, recommendations, and RAG systems.

  • Question answering
  • Text classification
  • Sentence similarity
  • Chinese
5.3K 441M–10M rows
Datasetopendatalab/ChartVerse-SFT-1.8MTop pick

Chart and diagram reasoning SFT dataset

1.8M chart QA pairs with Chain-of-Thought annotations for VLM training.

  • Visual question answering
  • Image text → text
  • English
5.1K 1411M–10M rows
Datasetlmms-lab/M4-Instruct-Data

M4 multi-image instruction sets

Multi-image instruction datasets for training multimodal models.

  • Visual question answering
  • Question answering
  • English
4.9K 78100K–1M rows
Datasetearakely-scale/CuaDriverBench-AgentEnvTop pick

AI agent benchmark for desktop multitasking with humans

Environment and execution logs for evaluating AI agents operating a computer alongside a human user.

  • Reinforcement learning
  • Image text → text
  • English
4.8K 2100K–1M rows
DatasetKingdrone-Junjue/EarthVLSet

Satellite imagery, VQA, and Earth surface segmentation

English geospatial dataset for multi-modal analysis and Earth imagery segmentation.

  • Visual question answering
  • Question answering
  • Image segmentation
  • English
4.8K 1210K–100K rows🔒 on request
Datasetnvidia/describe-anything-datasetTop pick

Detailed photo and video object descriptions from NVIDIA

English multimodal dataset for training models to describe specific regions in images and videos in detail.

  • Image → text
  • Video text → text
  • English
4.5K 60100K–1M rows
Datasetjordanparker6/publaynet

Layout markup for scientific article pages

Page images from PubMed Central with layout block annotations for document structure recognition.

  • Image → text
  • English
4.4K 4010K–100K rows
DatasetServiceNow/VideoCUATop pick

Annotated computer-use video recordings

55 hours of expert desktop app videos with manual annotations for training computer agents.

  • Video text → text
  • English
4.3K 3310K–100K rows
DatasetTeklia/IAM-line

English handwritten text lines for OCR

A classical dataset of individual handwritten text line images in English for OCR training.

  • Image → text
  • English
4.2K 3710K–100K rows
DatasetOpenDataArena/MMFineReason-Full-2.3M-Qwen3-VL-235B-ThinkingTop pick

2.3m reasoning examples from Qwen3-VL-235B

Raw corpus of 2.3 million multimodal tasks and reasoning chains (8.8 billion tokens) before quality filtering.

  • Visual question answering
  • Question answering
  • Text generation
  • English
4K 651M–10M rows
Datasetnvidia/Llama-Nemotron-VLM-Dataset-v1Top pick

Llama-Nemotron VLM Multimodal Dataset

NVIDIA dataset for training and evaluating the Llama-Nemotron VLM multimodal model.

  • Visual question answering
  • Image text → text
  • Image → text
3.9K 1731M–10M rows
DatasetTIGER-Lab/VisualWebInstruct

Multimodal instructions for VLM reasoning

Approximately 900K question-answer pairs (40% image-based, 60% text-based) for boosting multimodal model reasoning logic.

  • Question answering
  • Visual question answering
  • Image text → text
  • English
3.9K 451M–10M rows
Datasetbiglam/british-library-book-images

One Million 16th-19th Century Book Illustrations

Over 1 million illustrations extracted from digitized historical books from the British Library (1510–1900).

  • Image classification
  • Image → text
  • Text → image
3.8K 651M–10M rows
DatasetScaleAI/HSSTop pick

Intuitive Visual Reasoning Benchmark

A collection of images and short videos with questions from Scale AI to evaluate models' ability to understand hidden context.

  • Visual question answering
  • Video text → text
  • English
3.8K 19up to 1K rows
Datasetbevaya/ScreenSpot

Text-based UI element grounding benchmark

Over 1,200 instructions and screenshots from iOS, Android, macOS, Windows, and the web for evaluating multimodal models.

  • Text generation
  • Image → text
  • English
3.7K 531K–10K rows