DiSa: Saliency-Aware Foreground-Background Disentangled Framework for Open-Vocabulary Semantic Segmentation

TL;DR

DiSa introduces a saliency-aware framework for open-vocabulary semantic segmentation that effectively disentangles foreground and background features, leading to improved boundary localization and reduced bias.

Contribution

The paper proposes a novel saliency-aware disentanglement framework with hierarchical refinement to address foreground bias and spatial localization issues in VLM-based segmentation.

Findings

Outperforms state-of-the-art on six benchmarks

Effectively models foreground and background separately

Improves boundary accuracy and reduces bias

Abstract

Open-vocabulary semantic segmentation aims to assign labels to every pixel in an image based on text labels. Existing approaches typically utilize vision-language models (VLMs), such as CLIP, for dense prediction. However, VLMs, pre-trained on image-text pairs, are biased toward salient, object-centric regions and exhibit two critical limitations when adapted to segmentation: (i) Foreground Bias, which tends to ignore background regions, and (ii) Limited Spatial Localization, resulting in blurred object boundaries. To address these limitations, we introduce DiSa, a novel saliency-aware foreground-background disentangled framework. By explicitly incorporating saliency cues in our designed Saliency-aware Disentanglement Module (SDM), DiSa separately models foreground and background ensemble features in a divide-and-conquer manner. Additionally, we propose a Hierarchical Refinement Module (HRM) that leverages pixel-wise spatial contexts and enables channel-wise feature refinement through multi-level updates. Extensive experiments on six benchmarks demonstrate that DiSa consistently outperforms state-of-the-art methods.