Note: this works with Houdini 16.5. Other Houdini versions have not been confirmed by us yet, but it should work.
New Notes: For windows, this way of installing scipy does not function anymore with Houdini 17. Please follow this post: https://www.sidefx.com/forum/topic/58442/
To get older version of Houdini take a look at the daily builds page from SideFX.
Even Newer Notes: The newest version of Anaconda2 does not work even with version before Houdini 17. For windows users we now recommend installing Houdini 16.5 from the official distribution and installing Anaconda2-2.4.0-Windows-x86_64.exe from the official distribution. Completing the instructions of this post with with the above mentioned files should make this work.
MICROSOFT WINDOWS INSTALLATION GUIDE:
Window’s version of Houdini has its own’s version of python installed and does not use the default python location. Houdini uses python 2.7, NOT python 3.6. Make sure that you install the right scipy version. In anaconda for example, avoid the python 3.6 download.
Then go into the anaconda directory that is usually:
and copy all the files from there into the Houdini python directory
Scipy should now work fine.
Courtesy for the following Mac and Linux guides go to Chern et. al.
Mac Installation Guide
You don’t need to do anything. Mac already has SciPy.
Linux Installation Guide
Install SciPy from http://www.scipy.org/install.html. In particular it is recommended to install the Anaconda pack version 2.7. On Linux, Houdini ’s internal Python will read the Python library installed on your machine.
A great reason to use scipy will be the scipy sparse methods that speed up the matrix processing. A python node with scipy will be able to import scipy commands as seen in the code below with sparse.
from scipy.sparse import csr_matrix
import numpy as np
### get geometry
node = hou.pwd()
geo = node.geometry()
# continue with your normal python code