Skip to content

Inst-Inpaint

Remove objects from images with natural language prompts

Visit official site

What Inst-Inpaint is

The framework is described as an image inpainting algorithm that estimates the object to remove from natural language input and removes it at the same time. It includes the GQA-Inpaint dataset and the Inst-Inpaint framework, which removes objects from images based on text prompts.

What you can do with it

  • Remove a person from an image with a text prompt
  • Remove the background from an image with a text prompt
  • Remove an object from an image with a text prompt

What is included

  • The GQA-Inpaint dataset
  • The Inst-Inpaint inpainting framework
  • Various GAN and diffusion baselines
  • Different evaluation metrics for model quality and accuracy

Experiments

  • Experiments are run on synthetic and real image datasets
  • Evaluation metrics measure model quality and accuracy
  • The material describes significant quantitative and qualitative improvements

Pros & cons

✓ What we like

  • A text prompt names the object to remove
  • The dataset, framework and baselines are provided together

! What to watch out for

  • Object removal depends on estimating the target from natural language input
  • Experiments are run on synthetic and real image datasets

FAQ

How is an object removed?

The material describes entering a natural language request, such as asking to remove a person, a background or an object.

What is included with the method?

The material names the GQA-Inpaint dataset, the Inst-Inpaint framework, GAN and diffusion baselines and evaluation metrics.

Last reviewed: 2026-11-15

More AI image generation tools

View all →

How we review