Official Code for: Jinyuan Liu, Guanyao Wu, Junsheng Luan, Zhiying Jiang, Risheng Liu, Xin Fan*,“HoLoCo: Holistic and Local Contrastive Learning Network for Multi-exposure Image Fusion”, Information Fusion[J], 2023.
We strongly recommend that you use Conda as a package manager.
# create virtual environment
conda create -n holoco python=3.10
conda activate holoco
# select and install pytorch version yourself (Necessary & Important)
# install requirements package
pip install -r requirements.txtBefore testing or training HoLoCo, we strongly recommend downloading the following pre-trained model and placing them in ./checkpoints folder.
This code natively supports the same naming for over-/under-exposed image pairs. An naming example can be found in ./datasets/SICE folder.
# Test: use given example and save fused color images to result/SICE
# If you want to test the custom data, please modify the file path in 'test.py'
python start_test.py
# Train:
# Please prepare the custom data and change the modifiable options in 'start_train.py' (optional)
python start_train.pyIf this work has been helpful to you, we would appreciate it if you could cite our paper!
@article{liu2023holoco,
title={HoLoCo: Holistic and local contrastive learning network for multi-exposure image fusion},
author={Liu, Jinyuan and Wu, Guanyao and Luan, Junsheng and Jiang, Zhiying and Liu, Risheng and Fan, Xin},
journal={Information Fusion},
year={2023},
publisher={Elsevier}
}