EuroSciPy logo

EuroSciPy 2013

Brussels, Belgium - August 21-24 2013

Python for Image Procesing and Big Data

Fabrice Salvaire

Abstract

This talk is about the usage of Python for an industrial software project that aims to perform Image Processing on an automatized cover slip scanner platform in order to diagnose genetic diseases.

This five years experience demonstrates we could build an industrial software around Python and achieve the processing of big data, a digitised cover slip weights 80 GB of raw data, in no more than 10 min and to display such images in quasi real time.

The talk will give an overview and discuss the advantages and the actual shortcomings of Python for such a project:

  • CPU demanding and high level programming paradigm,
  • analysis steering and post-processing,
  • multi-core processing,
  • C/C++ binding,
  • software prototyping.

Moreover the talk will address these topics:

  • GUI,
  • GPU rendering,
  • Image Processing with Python,
  • Big Data storage,
  • Software Quality Assurance: large code maintenance, unit test, logging.

The Python ecosystem used by this project is:

  • Numpy
  • OpenCV, ITK, IPP
  • HDF5
  • zeromq
  • SqlAlchemy
  • Qt
  • OpenGL
  • SWIG
  • Scientific Linux OS

Sponsors