This repository is built in PyTorch 2.3.0, training and testing on Python3.9.15, CUDA12.4.
Follow these instructions:
-
Clone our repository
git clone https://github.com/OpenImagingLab/DualDn.git cd DualDn
-
Create and activate a new Conda environment
conda create -n pytorch_DualDn python=3.9.15 conda activate pytorch_DualDn
-
Install dependencies
pip install -r requirements.txt
-
Install ExifTool for extracting Exif metadata from raw images
Install ExifTool by following the official installation guide on how to install it on the Windows, MacOS or Unix platforms.
Below, fyi, some helpful tips are provided:
-
DualDn is tested with ExifTool version 12.84, so any version newer than 12.84 should work as well.
-
Check if ExifTool is correctly installed by running the following command:
exiftool -ver
If successful, it should return the version number such as
12.84
, if not, the installation is failed. -
On Windows platforms, ensure that ExifTool is added to system's GLOBAL PATH as well as the compiler's (e.g. in VSCode) PATH.
-
On Unix platforms (which are typically used as remote servers), after unzipping the
Image-ExifTool-12.84.tar.gz
file, you can add the following lines to '.bashrc' file for automated configuration:export PERL5LIB=/'the path to'/Image-ExifTool-12.84/install/share/perl5 export PATH=/'the path to'/Image-ExifTool-12.84/install/bin:$PATH
-