This repository serves as a comprehensive, curated resource hub for multispectral pedestrian detection. It targets researchers, engineers, and practitioners by consolidating essential datasets, state-of-the-art methods, annotation details, and evaluation tools. The primary benefit is accelerating development and comparative analysis in this specialized computer vision field by providing a centralized, organized collection of relevant materials.

How It Works

The resource functions as a structured index rather than an executable project. It meticulously lists and categorizes key multispectral pedestrian detection datasets, including KAIST, CVC-14, FLIR, LLVIP, and SMOD, detailing their characteristics, annotation counts, and acquisition specifics. A significant portion is dedicated to a chronological compilation of research papers from 2015 to the present, often linking directly to PDFs and associated code repositories, enabling users to explore and reproduce various detection methodologies. Additionally, it provides links to evaluation codes and annotation conversion tools.

Quick Start & Requirements

This repository is a curated list of resources and does not involve code execution or installation. No specific software, hardware, or dependencies are required to access or utilize the information provided.

Highlighted Details

• Extensive dataset catalog: Features prominent datasets like KAIST, CVC-14, FLIR, LLVIP, and SMOD, with details on image pairs, annotations, and acquisition conditions.
• Comprehensive methods timeline: Chronologically lists hundreds of research papers from 2015 onwards, many with direct links to PDFs and associated code repositories for deep learning approaches.
• Annotation clarity: Differentiates between various annotation versions for key datasets (e.g., KAIST), highlighting differences and their impact on evaluation.
• Tooling support: Includes links to evaluation codes and utilities for converting annotation formats (e.g., VBB to XML).

Maintenance & Community

As a static resource compilation, this repository does not feature active maintenance, community forums (like Discord/Slack), or ongoing development. It serves as a snapshot of available resources.

Licensing & Compatibility

No specific licensing information is provided for the resource itself. Users must refer to the individual licenses of the datasets, code repositories, and research papers linked within the repository for usage rights and compatibility.

Limitations & Caveats

The resource is a curated list and does not host the datasets or code directly, requiring users to follow external links. Some datasets, like FLIR, are noted to have initial alignment issues between modalities, necessitating the use of specific aligned versions. The quality and consistency of annotations can vary across different versions and datasets, requiring careful selection for specific research or application needs.