Installation & Setup

We highly recommand to use GPUs for the training. CUDA and MPI should be pre-installed in the device.

Pre-requirements

  1. nvidia gpu driver

  2. cudatoolkit 10.0 (for the current implemented tensorflow version)

_images/tensorflow_cuda_compatible.png

  1. openmpi

Linux or MacOS

Installation by command

Warning

The interface is an testing alpha version, we invite you for now to install from source. See below for installation from source.

Installation from source:

  1. go to https://github.com/ZeliangSu/SegmentPy and copy the HTTPS link

_images/download_from_git.png

and download by typing:

$ git clone https://github.com/ZeliangSu/SegmentPy.git
$ cd to/the/path/of/the/git/cloned/SegmentPy
$ python setup.py install

  1. or download the zip file and zip it, then:

    $ cd to/the/path/of/the/unzipped/SegmentPy
$ python setup.py install

Launch the SegmentPy by typing in the Terminal:

$ segmentpy

Frequent installation questions

Q: Can SegmentPy run on a Windows workstation?

A: Unfortunately, we could not at the current state garentee the performance on Windows machine. We recently gathered some unexpected behavior reports on Windows and are currently working on it.