Rethinking Alignment and Uniformity in Unsupervised Semantic Segmentation
This addresses the challenge of aligning low-level visual features with semantic representations without supervision, which is incremental as it builds on existing MI-based methods.
The paper tackles the problem of representation collapse in unsupervised image semantic segmentation by proposing a Semantic Attention Network (SAN) with a new module that dynamically generates pixel-wise and semantic features, achieving state-of-the-art performance on benchmarks and outperforming some pretrained methods.
Unsupervised image semantic segmentation(UISS) aims to match low-level visual features with semantic-level representations without outer supervision. In this paper, we address the critical properties from the view of feature alignments and feature uniformity for UISS models. We also make a comparison between UISS and image-wise representation learning. Based on the analysis, we argue that the existing MI-based methods in UISS suffer from representation collapse. By this, we proposed a robust network called Semantic Attention Network(SAN), in which a new module Semantic Attention(SEAT) is proposed to generate pixel-wise and semantic features dynamically. Experimental results on multiple semantic segmentation benchmarks show that our unsupervised segmentation framework specializes in catching semantic representations, which outperforms all the unpretrained and even several pretrained methods.