PaintSeg: Training-free Segmentation via Painting

TL;DR

PaintSeg introduces an innovative training-free, unsupervised segmentation method that leverages adversarial masked contrastive painting with iterative inpainting and outpainting to improve segmentation accuracy without supervision.

Contribution

The paper presents a novel training-free segmentation approach using adversarial masked contrastive painting with iterative inpainting and outpainting, eliminating the need for labeled data.

Findings

Outperforms existing methods in coarse mask, box, and point prompts

Works effectively with various prompt types without training

Provides a training-free, unsupervised segmentation solution

Abstract

The paper introduces PaintSeg, a new unsupervised method for segmenting objects without any training. We propose an adversarial masked contrastive painting (AMCP) process, which creates a contrast between the original image and a painted image in which a masked area is painted using off-the-shelf generative models. During the painting process, inpainting and outpainting are alternated, with the former masking the foreground and filling in the background, and the latter masking the background while recovering the missing part of the foreground object. Inpainting and outpainting, also referred to as I-step and O-step, allow our method to gradually advance the target segmentation mask toward the ground truth without supervision or training. PaintSeg can be configured to work with a variety of prompts, e.g. coarse masks, boxes, scribbles, and points. Our experimental results demonstrate that PaintSeg outperforms existing approaches in coarse mask-prompt, box-prompt, and point-prompt segmentation tasks, providing a training-free solution suitable for unsupervised segmentation.