Multi-Spectral Dataset

Dataset Overview

Dataset Name

Multi‑Spectral Dataset

Purpose

To evaluate multi‑spectral motion estimation methods, especially for visual navigation systems under complex lighting conditions.

Contents

• Sensor Configuration: Includes a standard RGB camera, an LWIR thermal camera, and a Kinect2.
  • The RGB and LWIR cameras provide 640×480 color and thermal images at 32 Hz.
  • Kinect2 supplies depth images at 30 Hz.
  • Ground‑truth poses from a motion‑capture system are recorded at 120 Hz.

Features

• Hardware‑synchronized multi‑spectral images: Ensures temporal alignment between color and thermal streams.
• Diverse sequences: Indoor and outdoor environments such as offices, stairs, hallways, elevators, roads, and buildings.
• Challenging lighting: Bright, dim, and complex illumination conditions are included.

Applications

• Benchmarking multi‑spectral SLAM and visual odometry algorithms.
• Testing and debugging visual navigation systems under varying lighting.

Tools

• Evaluation tools: MATLAB and Python scripts for assessing visual odometry or SLAM performance.
• Bag extractor: Utility to extract multi‑spectral data from ROS bags and generate dense depth maps.

Contact

• Contacts: Yu Zhang and Weichen Dai
• Email or issue tracker of the repository for inquiries.