Installation
It is highly recomended to create and use a new environment to install the plugin and its dependencies.
You can use mamba or conda to manage you environments but it's not necessary and you can use python venv
as well.
Setup
We provided install.sh
for Linux & Mac OS users, and install.bat
for Windows users.
First you need to clone the repo:
Step by Step
-
Create a new environment:
-
Activate the environment:
-
Install
This will install the appropriate PyTorch binaries without user intervention by automatically identifying compatible CUDA versions from the local setup.torch
andtorchvision
:
You can follow the instruction from here. But we can also uselight-the-torch
package: -
Installing all other dependencies:
This will install all dependencies includingnapari
,segment-anything
andsam-2
. -
Finally, install the plugin:
Requirements
python>=3.10
numpy<2.2
pytorch>=2.5.1
torchvision>=0.20.1
timm
segment-anything
sam-2
opencv-python
scikit-learn
scikit-image
scipy
matplotlib
pyqt
magicgui
qtpy
napari
h5py
pynrrd
pooch
There is also a pypi package available that you can install FF using pip
:
Note
Before install featureforest
using pip
you need to install segment-anything
and sam-2
manually.
For detailed napari installation see here.