Boosting Unsupervised Semantic Segmentation with Principal Mask Proposals

TL;DR

This paper introduces PriMaPs, a method leveraging large pre-trained models to improve unsupervised semantic segmentation by decomposing images into meaningful masks and fitting class prototypes, achieving competitive results.

Contribution

The paper proposes PriMaPs and PriMaPs-EM, a novel approach for unsupervised segmentation that enhances existing methods using principal mask proposals and a stochastic EM algorithm.

Findings

Competitive performance across multiple datasets and models

Boosts results when combined with existing segmentation pipelines

Effective with various pre-trained backbone models

Abstract

Unsupervised semantic segmentation aims to automatically partition images into semantically meaningful regions by identifying global semantic categories within an image corpus without any form of annotation. Building upon recent advances in self-supervised representation learning, we focus on how to leverage these large pre-trained models for the downstream task of unsupervised segmentation. We present PriMaPs - Principal Mask Proposals - decomposing images into semantically meaningful masks based on their feature representation. This allows us to realize unsupervised semantic segmentation by fitting class prototypes to PriMaPs with a stochastic expectation-maximization algorithm, PriMaPs-EM. Despite its conceptual simplicity, PriMaPs-EM leads to competitive results across various pre-trained backbone models, including DINO and DINOv2, and across different datasets, such as Cityscapes, COCO-Stuff, and Potsdam-3. Importantly, PriMaPs-EM is able to boost results when applied orthogonally to current state-of-the-art unsupervised semantic segmentation pipelines. Code is available at https://github.com/visinf/primaps.