FoCal

Toolkit for Evaluation, Fusion and Calibration of statistical pattern recognizers

At present the FoCal toolkit has two branches:

  • The original FoCal is applicable to any two-class recognizer and has been specialized for the task of speaker detection, as found in the NIST Speaker Recognition Evaluations . For more details on this branch, see FoCal Two-Class.
  • The multi-class FoCal branch is applicable to multi-class recognizers, with some specialization for the NIST Language Recogition Evaluations. Although some of the multi-class tools are generalizations of the two-class versions, the multi-class tools do not replace all of the functionality of the two-class tools. Users may therefore need both branches. For the multi-class branch see FoCal Multiclass.
The successor of the FoCal Toolkit is the BOSARIS Toolkit. It does not have multiclass capability, but for two classes it has improved and additional capabilities.