Jinyuan Liu, Zengxi Zhang, Jiahao Zhang, Zhiying Jiang, Long Ma, Xin Fan, Risheng Liu* ,“Universal Representation for Real-world Misaligned Infrared-Visible Image Fusion”, under review, 2025.
[2025-09-05] The proposed RIVIR dataset is available [2025-09-07] The testing code for RIVIR is available
Download the pre-trained model and put it in snapshot
Pytorch==1.12.1
python test_rivir.py
we introduce the first large-scale Real-world Infrared and Visible Image Registration Dataset (RIVIR) in the fifield, which contains a total of 1,300 pairs of unaligned image samples. The dataset is collected using a FLIR T1050sc dual� spectrum camera, with visible images captured at a resolution of 1280 × 960 and infrared images at 1024 × 768. To ensure diverse geographic distribution, data collection was conducted across multiple cities, spanning 26 degrees in longitude and 18 degrees in latitude.