This repository is a continuously updated, curated list of research papers, datasets, and code related to multi-modal object tracking (MMOT), specifically focusing on single-object tracking. It serves as a valuable resource for researchers and practitioners in computer vision aiming to stay abreast of the latest advancements in this specialized field.

How It Works

The project functions as a comprehensive survey and aggregation of MMOT research. It categorizes and lists relevant datasets, papers, and associated code, providing links to arXiv preprints, project pages, and GitHub repositories. This structured approach allows users to quickly navigate and access the latest developments across various modalities, including RGB, RGB-Depth (RGBD), RGB-Event (RGBE), RGB-Thermal Infrared (RGBT), and hyperspectral data.

Quick Start & Requirements

This repository is a curated list and does not have installation or execution requirements. It serves as a reference guide.

• Relevant Links:
  • Awesome Visual Object Tracking (VOT): https://github.com/Awesome-VOT/Awesome-VOT
  • Project Page for UW-COT220 and VL-SAM2: [project] (link not provided in README)
  • arXiv for UW-COT220 and VL-SAM2: [arXiv] (link not provided in README)
  • arXiv for VL-SOT500 and COST: [arXiv] (link not provided in README)

Highlighted Details

• Continuously updated with recent publications, including accepted papers for CVPR 2025 workshops and NeurIPS 2024.
• Covers a wide spectrum of multi-modal tracking, including vision-language, RGB-Event, RGB-Depth, RGBT, and hyperspectral tracking.
• Provides links to numerous datasets and associated code for benchmarking and reproduction.

Maintenance & Community

The project is maintained by 983632847/Awesome-Multimodal-Object-Tracking and welcomes contributions via pull requests. Suggestions can be sent to andyzhangchunhui@gmail.com.

Licensing & Compatibility

The project is released under the MIT license.

Limitations & Caveats

While comprehensive, the README does not provide direct links for all mentioned project pages or arXiv papers, requiring users to search for them manually. Some entries are dated as far back as 2012, indicating a long history of collected research.