git clone https://github.com/mrmdlab/Segmenthor.git
cd segmenthor
conda create -n segmenthor -c conda-forge python=3.11 nibabel pygame
conda activate segmenthor
pip3 install torch==2.0.1 torchvision==0.15.2 --index-url https://download.pytorch.org/whl/cpu
pip3 install opencv-python
git submodule update --init
pip3 install --no-index ./segment-anythingOn HPC, you need to enable the module mesa. If the following command doesn't work, consult your HPC administrator.
module load mesaDouble click start.ps1 to start Segmenthor. For the first time, a model file (357MB) will be downloaded. If you are using Windows and start.ps1 doesn't work, run the following command and try again.
Set-ExecutionPolicy RemoteSigned CurrentUserRead the tutorial for details.
While you can select slices you're interested in and compute their embedding on demand, you are advised to use precompute to compute the image embedding for all the slices in a single NIfTI file or even for all the anat images in a BIDS folder. It comes in handy if you have access to an HPC. Otherwise, you can leave your computer running overnight and all the image embedding will be ready the next morning.
# show help
precompute.ps1
https://mrmdlab.github.io/products/#Segmenthor
Disable Chinese input method, if any, before using this software. 使用前必须禁用中文输入法