Learning for Meta-Recognition
Walter J. Scheirer, Anderson Rocha, and Terrance E. Boult
[ IEEE T.IFS, V. 7, Issue 4, August 2012 ]
Abstract. In this work, we consider meta-recognition, an approach for post-recognition score analysis, whereby a prediction of matching accuracy is made from an examination of the tail of the scores produced by a recognition algorithm. This is a general approach that can be applied to any recognition algorithm producing distance or similarity scores. In practice, meta-recognition can be implemented in two different ways: a statistical fitting approach based on the Extreme Value Theory, and a machine learning approach utilizing features computed from the raw scores. While the statistical approach establishes a strong theoretical basis for meta-recognition, the machine learning approach is more accurate in its predictions. In this article, we present a study of the machine learning approach and its associated features for the purpose of building a highly accurate meta-recognition system for security and surveillance applications. Through the use of feature- and decision-level fusion, we achieve levels of accuracy well beyond those of the statistical approach, as well as the popular “cohort” model for post-recognition score analysis. In addition, we also explore the theoretical question of why machine learning based approaches tend to outperform statistical meta-recognition, and provide a partial explanation. We show that the introduced solutions are effective for a variety of different recognition applications across security and forensics-oriented computer vision, including biometrics, object recognition, and content-based image retrieval.
Index Terms. Meta-Recognition, Performance Modeling, Object Recognition, Face Recognition, Fingerprint Recognition, Content-Based Image Retrieval, Multi-biometric Fusion, Similarity Scores, Machine Learning
Click here to download our implementation for improving recognition systems using our proposed Machine Learning Meta-Recognition approach.