PyCVF Installation Procedure

From the 0.2 release, PyCVF is now modularized. So the installation is also modularized, is more standardized.

The core of PyCVF is now a pure python - package that relies on PyQT4 for its UI, and
on numpy, scipy, matplotlib for numeric computations, it uses BZR and launchpad

for the managing the project. So any system where these elements are installed can run PyCVF.

The other elements (OpenCV, Cuda....) dependencies are left to the packages.


Installing the necessary packages

The first thing to is to be sure that you have the necessary packages, on debian/ubuntu. Some of the packages are not absolutly mandatory but they are recommended:

## essential python
sudo apt-get install python-dev cython python-ctypes python-setuptools
# graphical interface
sudo apt-get install python-qt4  python-pyglet
# numpy and scipy install them, ipython
sudo apt-get install python-scientific python-numpy python-scipy python-matplotlib ipython
# PyCVF package system may help you installing necessary packages, provides it the necessary package to work
sudo apt-get install bzr python-launchpadlib python-apt python-launchpad-integration
# maybe not required by the core PyCVF, many other package will use also these famous package install them ::
sudo apt-get install python-beautifulsoup python-imaging
## this small package is also recommended
sudo easy_install termcolor

Install PyCVF

Then fetch the core system of PyCVF:

sudo easy_install pycvf

At the first run of a pycvf program, if the files that establish your identity and the fact that you accept the PyCVF license, are not found in your home directory, PyCVF will remind you the way it is licensed, and will ask you to provide information about your identity. This information will be stored in a local-file. At this moment, you may also decide to register as a PyCVF user, this is of course mandatory, but this allow to keep a track of our users. Unless you explicitly ask us to keep you informed, we will not use the fact you register as a ground to send you non-solicited emails.

So far we have only install a “minimal core” of pycvf, for working a certain number of basic packages are required:

sudo pycvf-admin install-package image video audio indexes rctree opencv indexes
sudo pycvf-admin build-packages
sudo pycvf-admin install-programs

You are invited to try more packages have a look at the directory below.

Check your system

Check your system works:


pycvf-admin install-dataset WANG
pycvf -d 'official_dataset("WANG")'

You may check your system status by typing:

pycvf-admin sysinfo

Start developping

Set up a good environment for working. IE, you need to have directories in your home directory that are search by Python and by your shell. Also, due to a limitation of Python, you have to actually mirror the “system pycvfext” directory, because from now your home direc:

cd ~

## check whether you have a directory for programs to be runned in the shell
test -d bin || mkdir bin

## check whether you have a directory for storing your python modules :
test -d python-lib || mkdir python-lib

## Now let's prepare our pycvfext directory
cd python-lib
pycvf-admin create-pycvfext-directory

## And let's create
pycvf-admin modify-bashrc -r $PWD/pycvfext
. ~/.bashrc

Start your own package, (let’s call it stereovision):

pycvf-admin start-package stereovision

If the path accessed by python is too long for you you are invited to create a symbolical link to your package

ln -s ~/python-lib/pycvfext/stereovision ~/stereovision

Then you are ready to develop, your own datasets, nodes, dataflows, scripts, experments, datatypes.

When you are satisfied you may publish your package:

pycvf-admin register-package stereovision


Installing the necessary packages

For the moment, the recommended version of Python to be used on windows is Python 2.6. As on Unix, the first thing to do is to be sure that you have the necessary packages , on Windows:

  • Install Python, setuptools, numpy, scipy, matplotlib, PIL: I recommend windows user to use a Python Distribution.
  • Install PyQT4
  • Install Mingw MSYS

Make accessible python from your environment:

ln -s /c/Python26/Scripts/* bin
Check easy_install works, and install the other essential package you may not have yet::
easy_install pyglet beautifulsoup bzr

Fetch a binary release of FFMPEG, Decompress it, fetch the dlls that are in the “bin” directory and copy them to your windows system directory.

Get the pyffmpeg binary package for your version of python.

Install PyCVF

Fetch the core system of PyCVF, in MSYS console type the following:

easy_install pycvf

At the first run of a pycvf program, if the files that establish your identity and the fact that you accept the PyCVF license, you will be required to accept them.

Install a first set of package:

pycvf-admin install-package image video audio indexes rctree
pycvf-admin build-packages
pycvf-admin install-programs

Other Unixes and MacOS

The installation must be quite similar to windows and linux description.

Mac OS X users will have to use Fink or MacPort to install the necessary packages before. I know PyQT4 takes lot of time to compile, but it is necessary...

Some Other Packages You May Want to Install


This package provides you a demo extensible website that allows you to browse the PyCVF datasets. And to do some of the operations online.


This package some 3d interfaces for various PyCVF tasks.

(Old) PyCVF 0.1 Installation Procedure


So forth, all installations are done in a way, were the user is invited to have its own branch of the code and to modify it. In further version, we may think about providing to the user some specific “project” branch like web-framework do it. For the moment it is not the case.


If you are under ubuntu, the easiest way to do the installation for the moment is to use the installation script :

This installation script, will check that you have the necessary packages on your system, retrieve branch of the project, modify your “.bashrc” to make your python envirionment ready. Invite you to make registration, and compile the additional wrappers for your system. If you are lucky everything shall go fine, else try to understand what’s going on, and if necessary open a ticket on the installation forum.

Hence, once the installation has been finished be sure to reload your bashrc:

. ~/.bashrc

Once you finished that you should be able to watch picture from a directory by doing:

pycvf_dbshow --db ''

Ubuntu/Debian packages shall be made available in near future from My Launchpad PPA . For the moment, there is no automated install for you. The install script may work on your distrib.

However, the best is to read the install-script, and to translate it for your distrib,

Other Unixes and MacOS

Dependint on your distribution, the simplest solution may be to install a virtualized ubuntu in your operating system by the mean of “virtual box”.

If you really want to build PyCVF for these platforms, for the moment you are on your own, but basically if you follow the instructions that are applied by the pycvf_install_on_ubuntu script, you should be able to get a working version of PyCVF on your mac.


The simplest solution may be to install a virtualized ubuntu in your operating system by the mean of “virtual box”.

For those that really require a native PyCVF for windows, we can confirm that we have already managed to make work PyCVF under Windows, but you need to be quite experienced with the porting of unix software to windows. Currently, there is no distribution of PyCVF for windows.