This repository serves as a comprehensive, curated collection of research papers and code related to Mamba models in computer vision. It aims to provide a centralized resource for researchers and practitioners exploring Mamba's potential as an alternative to traditional Transformer architectures for visual tasks. The collection highlights Mamba's efficiency and effectiveness in various vision applications, offering a valuable overview of the rapidly evolving field.

How It Works

The repository organizes papers by Mamba backbone networks, specific vision applications (e.g., image, remote sensing, medical imaging, video, point clouds), and other domains like reinforcement learning and graph learning. It includes links to papers, code repositories, and sometimes detailed performance comparisons, facilitating a deep dive into Mamba's capabilities and applications.

Quick Start & Requirements

This repository is a curated list of research papers and does not have a direct installation or execution command. Users are directed to individual paper repositories for specific code and setup instructions.

Highlighted Details

• Extensive coverage of Mamba backbone networks and their adaptations for vision tasks.
• Categorization of applications across diverse modalities including natural images, medical scans, videos, and point clouds.
• Inclusion of papers from top-tier conferences like NeurIPS, ICML, CVPR, ECCV, and MICCAI, indicating high-quality research.
• Regular updates with the latest research, as evidenced by recent news entries and paper additions.

Maintenance & Community

The repository is maintained by Rui Xu and colleagues from HKUST SMART Lab. It encourages community contributions and provides contact emails for questions.

Licensing & Compatibility

The repository itself is not software and thus not subject to software licensing. Individual linked code repositories will have their own licenses.

Limitations & Caveats

As a curated list of research, this repository does not provide runnable code or pre-trained models directly. Users must refer to individual paper links for implementation details and potential dependencies.