This repository contains an unofficial Pytorch implementation of the DeepFuse network for image fusion of extreme exposure image pairs, published in ICCV 2017.
This release of DeepFuse implementation was tested in Google Colab with
The dataset used for this experiment is the subset of the SICE dataset given here: https://github.com/ytZhang99/CF-Net
The SICE dataset can be found here: https://github.com/csjcai/SICE
Place ground truth, over exposed, and under exposed images for training in the following folder structure.
- SICE_subset
- test_data
- GT
- OE
- UE
- train_data
- GT
- OE
- UE
- val_data
- GT
- OE
- UE
python Main.py --train True --use_cuda True --trainset "./SICE_subset/train_data/"
Set the use_cuda
flag to False
if necessary.
python Main.py --train False --use_cuda True --testset "./SICE_subet/test_data/"
Thanks to Kede Ma et al. for their PyTorch implementation of the MEFSSIM metric from https://github.com/makedede/MEFNet