Framework for Research on Detection Classifiers

Detection, Face Detection, Classification, Image Processing, Computer Vision, AdaBoost, WaldBoost, Cascade of Classifiers, Corner Points, Classifier Evaluation

Detection of patterns in images with classifiers is currently one of the most important research topics in computer vision. Many practical applications such as face detection exist and recent work even suggests that any specialized detectors (e.g. corner-point detectors) can be approximated by very fast detection classifiers. In this paper, we analyze the requirements on tools which are needed when experimenting with detection classifiers and we present a general framework which was created to fulfill these requirements. This framework offers high performance for training, high variability, elegant handling of configuration and it is able to meet all the requirements which arise when experimenting with almost all possible kinds of detection classifiers. The framework offers good testing support, full supporting infrastructure and some useful training algorithms and features. We offer this framework for research and educational purposes and we hope it will allow lower initial investments when experimenting with detection classifiers.