Skip to content

Dataset · paint-by-inpaint

Paired images and instructions for object addition

paint-by-inpaint/PIPE

Around one million "before/after" image pairs with text instructions for mask-free image editing.

What's inside

A large multimodal dataset (image-to-image and text) containing about 1 million image pairs (1M to 10M in metadata). Each record contains an original image, a target image, and a natural language text instruction describing the addition of a new object. Created specifically to advance mask-free instruction-based image editing models.

What you can do with it

  • Training and fine-tuning image-to-image diffusion models for text-prompt photo editing
  • Training LoRA adapters for cleanly adding objects to a scene without manual mask selection
Tasks
Image → image
Size
1M–10M rows
Downloads this month
8K
Likes
33

How to load it

from datasets import load_dataset

ds = load_dataset("paint-by-inpaint/PIPE")