Skip to content

Dataset · CaptionEmporium

57.3M multimodal descriptions for the Megalith-10m base

CaptionEmporium/flickr-megalith-10m-internvl2-multi-caption

A massive corpus of 57.3M short and long English captions for 10M Flickr images from various VLM models.

What's inside

It contains about 57.3 million synthetic English text descriptions for images from the madebyollin/megalith-10m set (repository size ranges from 1 to 10 million rows in tabular and text formats). Long and short captions were generated for each image simultaneously by several models: InternVL2 8B, Florence2, and ShareCaptioner.

What you can do with it

  • Training and fine-tuning text-to-image generators and LoRA on detailed captions
  • Training image-to-text description generation models
Tasks
Text → image, Image → text, Other
Languages
English
Size
1M–10M rows
Downloads this month
548
Likes
33

How to load it

from datasets import load_dataset

ds = load_dataset("CaptionEmporium/flickr-megalith-10m-internvl2-multi-caption")