DiffV2IR introduces a groundbreaking diffusion model that translates visible images to infrared with semantic awareness and wavelength diversity.

• Overcomes three major challenges: semantic-aware translation, infrared wavelength diversity, and limited datasets
• Leverages vision-language understanding to preserve crucial semantic details
• Adapts to various infrared wavelength bands for versatile security applications
• Demonstrates superior performance despite scarce training data

This research significantly advances security imaging technology for surveillance, threat detection, and nighttime monitoring, offering more reliable visual information in challenging visibility conditions.

DiffV2IR: Visible-to-Infrared Diffusion Model via Vision-Language Understanding