Joint Demosaicing and Deghosting of Time-Varying Exposures for Single-Shot HDR Imaging

Project page | Paper | Supplemental

Jungwoo Kim, Min H. Kim KAIST

Tested Environments

• OS: Ubuntu 16.04 / Windows 10
• Graphic card: TITAN RTX / RTX 2060
• Cuda toolkit version: 10.2
• NVIDIA Driver Version: 456.71
• python: 3.7
• torch version: 1.9.0

Installation

This repository is built in Pytorch 1.9.0 and tested on Ubuntu 16.04 enviornment (Python3.7, CUDA10.2, cuDNN7.6).

Please follow below instructions:

1. Clone our repository

git clone git@github.com:KAIST-VCLAB/singshot-hdr-demosaicing.git
cd singshot-hdr-demosaicing

2. Make conda enviornment

conda create -n pytorch190 python=3.7
conda activate pytorch190

3. Install dependencies

conda install pytorch==1.9.0 torchvision==0.10.0 torchaudio==0.9.0 cudatoolkit=10.2 -c pytorch
pip install -r "requirements.txt"

Demo

First download our pretrained model and put best_psnr_mu.pt file in code/models/best_psnr_mu.pt.

To run demo with pre-trained models, run below code:

cd code
python test.py