Meta-Recognition Project

Welcome to the Meta-Recognition project webpage. Here you will find solutions and tools for improving recognition systems including downloadable library (libMR) which is free for non-commercial use.

Here is what you can do with our solutions:

  • Attribute Fusion – robust fusion for descriptive visual attributes in a binary         SVM context for image retrieval applications.                                                                 For more information click Attribute Fusion
  • EVT Failure prediction – a performance prediction method for recognition algorithms based on the use of the statistical extreme value theory (EVT).
    For more information click EVT-MR
  • ML Failure prediction – a performance prediction method for recognition algorithms based on the use of machine learning classifiers
    For more information click ML-MR
  • EVT and ML Fusion – strategies for fusing recognition algorithms based either on statistical extreme value theory or machine learning theory
    For more information click EVT-MRML-MR, or EVT-FUSION
  • OpenSet Recognition and 1-vs-Set Machine – Recognition solutions to deal with the Open Set nature of recognition problems. For more information click Open Set
  • Exemplar Codes  combining an exemplar-SVM like approach with EVT normalization to form a secondary feature vector for recognition systems. For more information click ExemplarCodes