# Inpaint Anything: Segment Anything Meets Image Inpainting

**Authors:** Tao Yu, Runseng Feng, Ruoyu Feng, Jinming Liu, Xin Jin, Wenjun Zeng, Zhibo Chen  
**arXiv:** [2304.06790](https://arxiv.org/abs/2304.06790)  
**Date:** 2023-04-13  
**Summary page:** https://www.summarizepaper.com/en/paper/2304.06790v1/

## Key points

- Inpainting is a technique to fill in missing or damaged parts of an image
- Modern systems still struggle with mask selection and holes filling
- Inpaint Anything (IA) proposes a new paradigm of "clicking and filling" to solve inpainting-related problems
- IA combines the strengths of different models, including the Segment-Anything Model (SAM)
- SAM allows users to remove anything from an image by clicking on it, and IA removes the object and smooths out the hole with context
- Users can provide text-based prompts to fill the hole with corresponding generative content via driving AIGC models like Stable Diffusion
- IA supports three main features: Remove Anything, Fill Anything, and Replace Anything
- The proposed method supports mask-free image inpainting using SAM, making it easier for users to select objects without creating masks manually
- The authors provide their codes at https://github.com/geekyutao/Inpaint-Anything for everyone interested in sharing and promoting new projects based on their work.
- Experiments conducted by the authors demonstrate its potential as a valuable tool for researchers working in computer vision and related fields.

## Summary

Inpainting is a technique used to fill in missing or damaged parts of an image. Modern image inpainting systems have made significant progress but still struggle with mask selection and holes filling. Tao Yu and his team propose a new paradigm of "clicking and filling" called Inpaint Anything (IA) that combines the strengths of different models to build a powerful and user-friendly pipeline for solving inpainting-related problems. The core idea behind IA is based on the Segment-Anything Model (SAM), which allows users to remove anything from an image by clicking on it. IA then removes the object and smooths out the hole with context. Users can also provide text-based prompts to IA, which will fill the hole with corresponding generative content via driving AIGC models like Stable Diffusion. With IA, users have another option to retain the click-selected object and replace the remaining background with newly generated scenes. This paper presents the first attempt at mask-free image inpainting using SAM, making it easier for users to select objects without having to create masks manually. The proposed method supports three main features: Remove Anything, Fill Anything, and Replace Anything. The authors also provide their codes at https://github.com/geekyutao/Inpaint-Anything for everyone interested in sharing and promoting new projects based on their work. Experiments conducted by the authors demonstrate its potential as a valuable tool for researchers working in computer vision and related fields.