The release 1.2.0 of the Bob signal-processing and machine learning toolbox is now available

Bob provides both efficient implementations of several machine learning algorithms as well as a framework to help researchers to publish reproducible research.
It is developed by the Biometrics Group at Idiap in Switzerland.

The previous release of Bob was providing:
Gaussian Mixture Models (GMMs), Bayesian intra/extra (personal) classifier, Inter-Session Variability modeling (ISV), Joint Factor Analysis (JFA), 
Probabilistic Linear Discriminant Analysis (PLDA).

The new release of Bob has brought the following features and/or improvements, such as:
  • Unified implementation of Local Binary Patterns (LBPs),
  • Histograms of Oriented Gradients (HOG) implementation,
  • Total variability (i-vector) implementation,
  • Conjugate gradient based-implementation for logistic regression,
  • Improved multi-layer perceptrons implementation (Back-propagation can now be easily used in combination with any optimizer -- i.e L-BFGS),
  • Pseudo-inverse-based method for Linear Discriminant Analysis,
  • Covariance-based method for Principal Component Analysis,
  • Whitening and within-class covariance normalization techniques,
  • Module for object detection and keypoint localization (bob.visioner),
  • Improved extensions (satellite packages), that now support both Python and C++ code, within an easy to use framework,
  • Improved documentation and add new tutorials,
  • Support for Intel's MKL (in addition to ATLAS),
  • Extend supported platforms (Arch Linux).

This release represents a major milestone in Bob with plenty of functionality improvements (>640 commits in total: ) and plenty of bug fixes ( ).
  • Ubuntu: 10.04, 12.04, 12.10 and 13.04
  • For Mac OSX: works with 10.6 (Snow Leopard), 10.7 (Lion) and 10.8 (Mountain Lion)

For instructions on how to install pre-packaged version on Ubuntu or OSX, consult our quick installation instructions: (N.B. OS X macport has not yet been upgraded. This will be done very soon. cf. ).