Haozhen Wei, Chengjun Jiang, Xingyuan Li, Jinyuan Liu, "MixStyle-augmented Meta-Learning for Cross-Domain Infrared-Visible Image Fusion", IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2026.
@inproceedings{wei2026mixstyle,
title={MixStyle-augmented Meta-Learning for Cross-Domain Infrared-Visible Image Fusion},
author={Wei, Haozhen and Jiang, Chengjun and Li, Xingyuan and Liu, Jinyuan},
booktitle={ICASSP 2026-2026 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)},
pages={9542--9546},
year={2026},
organization={IEEE}
}
# create virtual environment
conda create -n MixMetaFusion python=3.8
conda activate MixMetaFusion
# install requirements
pip install -r requirements.txt
Our checkpoints can be found in "ckpt/MixMetaFusion.pth". Then, you can test our pure fusion method through
python test.py --ckpt ckpt/MixMetaFusion.pth --datasets M3FD
python train.py --domain-roots M3FD=datasets/M3FD/train Rain=datasets/Rain/train Snow=datasets/Snow/train --epochs 180
You can color the output gray images for task-guided image fusion training and testing through
python tocolor.py --gray-folder datasets/M3FD/test/images --vi-folder datasets/M3FD/test/vi --out-folder output_color