English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

This paper introduces ProLab, a novel approach using property-level labelspace for creating strong interpretable segmentation models. Instead of relyingsolely on category-specific annotations, ProLab uses descriptive propertiesgrounded in common sense knowledge for supervising segmentation models. It isbased on two core designs. First, we employ Large Language Models (LLMs) andcarefully crafted prompts to generate descriptions of all involved categoriesthat carry meaningful common sense knowledge and follow a structured format.Second, we introduce a description embedding model preserving semanticcorrelation across descriptions and then cluster them into a set of descriptiveproperties (e.g., 256) using K-Means. These properties are based oninterpretable common sense knowledge consistent with theories of humanrecognition. We empirically show that our approach makes segmentation modelsperform stronger on five classic benchmarks (e.g., ADE20K, COCO-Stuff, PascalContext, Cityscapes, and BDD). Our method also shows better scalability withextended training steps than category-level supervision. Our interpretablesegmentation framework also emerges with the generalization ability to segmentout-of-domain or unknown categories using only in-domain descriptiveproperties. Code is available at https://github.com/lambert-x/ProLab.

A Semantic Space is Worth 256 Language Descriptions: Make Stronger Segmentation Models with Descriptive Properties