Skip to content

Dataset · jackyhate

2M text-image pairs for T2I

jackyhate/text-to-image-2M

A high-quality and diverse dataset of approximately 2 million text-image pairs for text-to-image training.

What's inside

text-to-image-2M is a dataset for text-to-image training containing high-quality and diverse text-image pairs based on its name and card. Metadata: en language, 1M-10M size. The card references a paper on aesthetic image generation with composition transfer.

What you can do with it

  • Train a diffusion T2I model
  • Fine-tune LoRA for style or composition
  • Evaluate composition transfer
Tasks
Text → image, Image → text, Image classification
Languages
English
Size
1M–10M rows
Downloads this month
7.2K
Likes
198

Models trained on it

How to load it

from datasets import load_dataset

ds = load_dataset("jackyhate/text-to-image-2M")