Skip to content

Latest commit

 

History

History
67 lines (47 loc) · 2.95 KB

installation.md

File metadata and controls

67 lines (47 loc) · 2.95 KB
Raw

Notes on software installation for NCams

FLIR Cameras Setup (Spinnaker)

  1. Install SpinView for testing whether the cameras work with full SDK.
  2. Test whether the cameras are properly setup by running SpinView (for example, located in C:\Program Files\Point Grey Research\Spinnaker\bin64\vs2015\SpinView_WPF.exe)
  3. Install Spinnaker module into your Python (after installing DeepLabCut, using pip install).

Do NOT try to install PySpin using pip, it will install a different, unrelated package.

DeepLabCut

DeepLabCut can be installed EITHER through an Anaconda approach or through system-wide installation. Anaconda approach is easier, while system-wide approach provides better flexibility and control.

  1. Install latest NVIDIA drivers. No need to install NVIDIA Experience.

  2. Install CUDA 10.0. When installing, choose a custom installation and DO NOT install other drivers (e.g. videocard drivers).

Continue either with Anaconda or system approach.

Setting up DLC using Anaconda approach

  1. Install Anaconda3

  2. Follow instructions on DeepLabCut README to install DeepLabCut.

Additional notes:

Setting up DLC using system-wide approach

  1. Install CuDNN 7.6.5 for CUDA 10.0

  2. Install Python 3.6. 3.6 needs to be used because DeepLabCut depends on tables module version 3.4.3, which we had problems with when compiling on Python 3.7.

  3. Install TensorFlow:

> pip install tensorflow-gpu=1.14

See below for instructions if you have more than one Python3 installed.

  1. Install the needed modules on Python:
    • numpy
    • matplotlib
    • scipy
    • DeepLabCut
    • Spyder
    • moviepy
    • opencv-contrib-python
    • reportlab
    • pyyaml
    • easygui The easiest way to install is to use pip from command line:
> pip install deeplabcut

If you have more than one Python installed (e.g. Python 3.7 and Python 3.6), you can target a specific one using py -<version> modifier, if you have py configured. You can target scripts like pip from them: py -3.6 -m pip. For example, this will install the module:

> py -3.6 -m pip install deeplabcut

If the installation script brakes at any point, check which module it could not install, then try a wheel for that module. The wheel is installed using pip from command line, e.g.:

Downloads> py -3.6 -m pip install Shapely-1.6.4.post2-cp36-cp36m-win_amd64.whl