This repository is an "Awesome List" curating attention mechanisms and plug-and-play modules for computer vision tasks. It serves researchers and practitioners by cataloging papers, their publication venues, and associated code repositories, facilitating the exploration and adoption of state-of-the-art attention techniques.

How It Works

The list is organized into categories: Attention Mechanism, Dynamic Networks, Plug and Play Modules, and Vision Transformers. Each entry links to the relevant research paper and, where available, its implementation. This structured approach allows users to quickly find and compare different attention architectures and their applications in computer vision.

Quick Start & Requirements

This is a curated list, not a runnable library. To use any of the included modules, users must refer to the individual project repositories linked within the list.

Highlighted Details

• Extensive catalog of attention mechanisms, including Squeeze-and-Excitation (SE), Convolutional Block Attention Module (CBAM), Non-local Networks, and various Transformer-based approaches.
• Covers "plug-and-play" modules and dynamic network concepts that can be integrated into existing CNN architectures.
• Features a dedicated section on Vision Transformers, highlighting key models like ViT, Swin Transformer, and MobileViT.
• Includes links to papers, GitHub repositories, and sometimes blog posts or specific implementations for each listed item.

Maintenance & Community

The list is maintained by pprp and welcomes community contributions via issues and pull requests for additions or corrections.

Licensing & Compatibility

The repository itself is licensed under an unspecified license. The licensing of individual code repositories linked within the list will vary and must be checked on a per-project basis.

Limitations & Caveats

The list is not exhaustive due to the maintainer's "limited ability and energy." Some entries may lack direct code links or comprehensive descriptions. The "Main Idea" column for Vision Transformers is often empty.