Improving object detection with boosted histograms

摘要

We address the problem of visual object class recognition and localization in natural images. Building upon recent progress in the field we show how histogram-based image descriptors can be combined with a boosting classifier to provide a state of the art object detector. Among the improvements we introduce a weak learner for multi-valued histogram features and show how to overcome problems of limited training sets. We also analyze different choices of image features and address computational aspects of the method. Validation of the method on recent benchmarks for object recognition shows its superior performance. In particular, using a single set of parameters our approach outperforms all the methods reported in VOC05 Challenge for seven out of eight detection tasks and four object classes while providing close to real-time performance.