Visible and Thermal Infrared Image Pairs for Unexploded Ordnance Detection

A collection of 234 sets of paired visible light and thermal infrared images for detecting unexploded ordnance. The data was gathered by JLU JRI-UXO in a single region under varying conditions including snow and tree branches. It was published in April 2026.

Use Cases

• Train a multimodal object detection model using paired visible and thermal infrared images to identify UXOs.
• Benchmark the performance of image fusion algorithms for target recognition in snowy environments.
• Evaluate model robustness by testing on data collected at different times of day, leveraging thermal infrared features.
• Develop domain adaptation techniques for object detection in scenes with tree branch occlusions.

Strengths

• Contains 234 distinct image sets, providing a foundation for model training.
• Includes paired visible and thermal infrared modalities for each scene.
• Covers specific environmental conditions like snow and sparse targets.

Limitations

• Limited to 234 image sets, which is a small sample size for deep learning.
• Geographic scope is restricted to a single, unspecified region.
• No provided annotations or bounding boxes for the targets.

Provenance

Source

JLU JRI-UXO

Collection Method

Field collection of visible light and thermal infrared images.

Time Range

Images were collected at different times, but specific dates are unknown.

Freshness

Last updated in April 2026.

Geography

First region (specific location unknown); scenes include snow and tree-covered areas.