From d581280a7fa47ad8419acccb14fd7048f00e470d Mon Sep 17 00:00:00 2001 From: Amir MOHAMMADI <amir.mohammadi@idiap.ch> Date: Tue, 19 Apr 2016 18:00:46 +0200 Subject: [PATCH] Explain about fusion implementation and how to run experiments. --- doc/fusion.rst | 25 +++++++++++++++++++++++++ 1 file changed, 25 insertions(+) create mode 100644 doc/fusion.rst diff --git a/doc/fusion.rst b/doc/fusion.rst new file mode 100644 index 0000000..a3c2b76 --- /dev/null +++ b/doc/fusion.rst @@ -0,0 +1,25 @@ +.. vim: set fileencoding=utf-8 : +.. author: Amir Mohammadi <amir.mohammadi@idiap.ch> + +.. _bob.fusion.base.fusion: + + +========================================================== +Running Score Fusion in Biometric Recognition Experiments +========================================================== + +There are two scripts available in the package as of now and they work for verification scenarios. +Spoofing and Anti-spoofing scenarios are not considered yet in the scripts however the Python API allows for easy extension. + +Each score fusion experiment requires at least a classifier. +The ``bob.fusion.base`` package itself implements two such classifiers: ``MLP`` and ``WeightedSum``. +You can also use any class as a classifier that implements a ``fit(X[, y])`` and a ``decision_function(X)`` method. +An example is `sklearn.linear_model.LogisticRegression <http://scikit-learn.org/stable/modules/generated/sklearn.linear_model.LogisticRegression.html>`_. + +You can also use pre-processors to pre-process your data. The pre-processor class should implement a ``fit_transform(X[, y])`` and a ``transform(X[, y, copy])`` methods. An example is `sklearn.preprocessing.StandardScaler <http://scikit-learn.org/stable/modules/generated/sklearn.preprocessing.StandardScaler.html>`_. + +Using ``sklearn`` classifiers and pre-processors you can different fusion algorithms. Please take a look at some examples in ``bob.fusion.base.config.algorithm`` to see how it is done. + +Also take a look at the scripts ``fuse.py`` and ``plot_fusion_decision_boundary.py`` and use them for your actual fusion experiments. + +.. include:: links.rst -- GitLab