A statistical modelling toolkit, providing all the tools required to build, fit, visualize, and test deformable models.
jabooth Merge pull request #815 from jabooth/pynormals
Pure-Python mesh normal calculation fallback
Latest commit 51471f8 Aug 1, 2018

README.md

menpo

Coverage Status PyPI Release BSD License
Python 2.7 Support Python 3.5 Support Python 3.6 Support

Menpo. The Menpo Project Python package for handling annotated data.

What is Menpo?

Menpo is a Menpo Project package designed from the ground up to make importing, manipulating and visualizing image and mesh data as simple as possible. In particular, we focus on annotated data which is common within the fields of Machine Learning and Computer Vision. All core types are Landmarkable and visualizing these landmarks is very simple. Since landmarks are first class citizens within Menpo, it makes tasks like masking images, cropping images inside landmarks and aligning images very simple.

Menpo were facial armours which covered all or part of the face and provided a way to secure the top-heavy kabuto (helmet). The Shinobi-no-o (chin cord) of the kabuto would be tied under the chin of the menpo. There were small hooks called ori-kugi or posts called odome located on various places to help secure the kabuto's chin cord.

--- Wikipedia, Menpo

Installation

Here in the Menpo Team, we are firm believers in making installation as simple as possible. Unfortunately, we are a complex project that relies on satisfying a number of complex 3rd party library dependencies. The default Python packing environment does not make this an easy task. Therefore, we evangelise the use of the conda ecosystem, provided by Anaconda. In order to make things as simple as possible, we suggest that you use conda too! To try and persuade you, go to the Menpo website to find installation instructions for all major platforms.

If you feel strongly about using Menpo with the most commonly used Python package management system, pip, then you should be able to install Menpo as follows:

> pip install cython numpy
> pip install menpo

However, this may be difficult to achieve on platforms such as Windows where a compiler would need to be correctly configured. Therefore, we strongly advocate the use of conda which does not require compilation for installing Menpo (or Numpy, SciPy or Matplotlib for that matter). Installation via conda is as simple as

> conda install -c menpo menpo

Build Status

And has the added benefit of installing a number of commonly used scientific packages such as SciPy and Numpy as Menpo also makes use of these packages.

CI Host OS Build Status
Travis Ubuntu 12.04 (x64) Travis Build Status
Jenkins OSX 10.10 (x64) and Windows 10 (x86, x64) Jenkins Build Status

Usage

Menpo makes extensive use of Jupyter Notebooks to explain functionality of the package. These Notebooks are hosted in the menpo/menpo-notebooks repository. We strongly suggest that after installation you:

  1. Download the latest version of the notebooks
  2. Conda install Jupyter notebook and IPython: conda install jupyter ipython notebook
  3. Run jupyter notebook
  4. Play around with the notebooks.

Want to get a feel for Menpo without installing anything? You can browse the notebooks straight from the menpo website.

Other Menpo projects

Menpo is designed to be a core library for implementing algorithms within the Machine Learning and Computer Vision fields. For example, we have developed a number of more specific libraries that rely on the core components of Menpo:

  • menpofit: Implementations of state-of-the-art deformable modelling algorithms including Active Appearance Models, Constrained Local Models and the Supervised Descent Method.
  • menpo3d: Useful tools for handling 3D mesh data including visualization and an OpenGL rasterizer. The requirements of this package are complex and really benefit from the use of conda!
  • menpodetect: A package that wraps existing sources of object detection. The core project is under a BSD license, but since other projects are wrapped, they may not be compatible with this BSD license. Therefore, we urge caution be taken when interacting with this library for non-academic purposes.
  • menpowidgets: This package provides high level object viewing classes using Matplotlib and Jupyter. Jupyter notebooks are therefore required to this package - and Menpo also implicitly relies on menpowidgets for any widget functionality.

Documentation

See our documentation on ReadTheDocs

Testing

We use nose for unit tests.

After installing nose and mock, running

>> nosetests .

from the top of the repository will run all of the unit tests.

Some small parts of Menpo are only available if the user has some optional dependency installed. These are:

  • 3D viewing methods, only available if menpo3d is installed
  • menpo.feature.dsift only available if cyvlfeat is installed
  • Widgets are only available if menpowidgets is installed