This repository serves as a comprehensive, curated list of publications and resources related to Open-Vocabulary Semantic Segmentation (OVSS) and its sub-fields, including zero-shot segmentation, referring image segmentation, and open-vocabulary object detection. It is a valuable resource for researchers and practitioners in computer vision aiming to develop models that can segment and understand arbitrary visual concepts described by natural language.

How It Works

The repository organizes research papers by their methodology (fully-supervised, weakly-supervised, training-free) and specific task within the OVSS domain. It provides links to papers, code repositories, and sometimes highlights key contributions or findings, facilitating a structured overview of the rapidly evolving field. The primary approach is to catalog and categorize existing work, enabling users to quickly find relevant research and understand the state-of-the-art.

Quick Start & Requirements

This is a curated list of research papers and does not involve direct installation or execution of code. Users can browse the categorized links to access research papers and their associated code repositories.

Highlighted Details

• Extensive coverage of Open-Vocabulary Semantic Segmentation, Zero-Shot Semantic Segmentation, Referring Image Segmentation, and Open-Vocabulary Object Detection.
• Includes papers from major computer vision conferences (CVPR, ICCV, ECCV, NeurIPS, ICLR, AAAI) and journals.
• Categorization by supervision type (fully-supervised, weakly-supervised, training-free) and task variations.
• Links to code repositories are provided for many entries, enabling practical exploration.

Maintenance & Community

The repository is actively maintained, with recent updates including papers from CVPR'25 and ArXiv'25. Suggestions for missing papers can be sent to tbh3223@mail.ustc.edu.cn or lydyc@mail.ustc.edu.cn.

Licensing & Compatibility

This repository is a list of links to external research papers and code. The licensing and compatibility of individual papers and code repositories are determined by their respective authors and projects.

Limitations & Caveats

As a curated list, the repository itself does not provide implementations or benchmarks. The quality and availability of linked code repositories can vary. The field is rapidly advancing, so the list may not be exhaustive of the very latest pre-prints.