Skip to content

Latest commit

 

History

History
63 lines (43 loc) · 2.5 KB

INSTALL.rst

File metadata and controls

63 lines (43 loc) · 2.5 KB
Raw

Installation

As noted in README.rst, it's much easier to use binary distributions of py-sdm through conda. See the instructions there.

If you don't want to or can't do that, you can install like any other python package, once you have the requirements set up properly.

Requirements

  • numpy and scipy are not included in the pip requirements, to avoid over-eager reinstallation with pip install -U.
  • FLANN is used for fast nearest-neighbor searches, through the cyflann library. You need a more recent version than 1.8.4, which currently means you need to install from source (or use my binary on binstar). A version with OpenMP support is assumed (which means you need to compile with gcc, not clang or llvm-gcc). Note, though, that on OSX, versions of gcc prior to 4.6 will cause segfaults when FLANN uses OpenMP parallelism nested between py-sdm code and FLANN's. I use gcc-4.8 from Homebrew.
  • vlfeat-ctypes, a minimal ctypes interface to the vlfeat computer vision algorithm library. This is installed by pip automatically, but make sure to run python -m vlfeat.download to download the library binary if you want to extract image features.

If you're using my conda channel but want to install py-sdm from source:

conda install distribute numpy scipy cython flann cyflann scikit-learn \
              scikit-image vlfeat-ctypes h5py nose progressbar

Actual installation

You have several options. The most straightforward way to get a snapshot is to just pip install the code (either into your system site-packages or a virtualenv) via:

pip install 'https://github.com/dougalsutherland/py-sdm/tarball/master#egg=sdm-0.1.0dev'

Since this package is still in early development, however, this somewhat complicates the process of upgrading the code. It's easier to get upgrades if you do a development install. If you have a checkout of the code, just do:

python setup.py develop

Pip wil also check out the code for you into (by default) ./src/py-sdm:

pip install -e 'git+https://github.com/dougalsutherland/py-sdm.git#egg=sdm-0.1.0dev'

In either case, if you issue a git pull in the source directory it'll update the code.