This is the approximate implementation of the DRHT paper. The project page can be found here:
https://ybsong00.github.io/cvpr18_imgcorrect/index.html
The ldr2hdr model is from HDRCNN http://hdrv.org/hdrcnn/. We will release our original implementation of ldr2hdr soon. The model can be found at either the HDRCNN project page or here:
https://drive.google.com/open?id=138JfKA5QzjDu78PLf6Ih9t5Qh2bBMvhI.
You need to download it at first and put it under checkpoint folder.The illustration of the files and folders.
############### folders ################
checkpoint --- pre-trained models
input --- input ldr images
hdr_output --- hdr files
samples --- ldr results
############### .py files ################
ldr2hdr.py and hdr2ldr.py define the ldr2hdr and hdr2ldr networks, respectively.
ldr2hdr_test.py and hdr2ldr_test.py provide simple evaluation.
############### notes ################
1. The ldr2hdr part is based on the Siggraph Asia 17 paper "HDR image reconstruction from a single exposure using deep CNNs".
2. The hdr2ldr part performs better when using large batch_size.
If you find the code useful, please cite the following papers:
@inproceedings{yang-cvpr18-DRHT,
author = {Yang, Xin and Xu, Ke and Song, Yibing and Zhang, Qiang and Wei, Xiaopeng and Rynson, Lau},
title = {Image Correction via Deep Reciprocating HDR Transformation},
booktitle = {IEEE Conference on Computer Vision and Pattern Recognition},
year = {2018},
}
@article{EKDMU17,
author = "Eilertsen, Gabriel and Kronander, Joel, and Denes, Gyorgy and Mantiuk, Rafa\l and Unger, Jonas",
title = "HDR image reconstruction from a single exposure using deep CNNs",
journal = "ACM Transactions on Graphics (TOG)",
year = "2017",
}