k-fold Subsampling based Sequential Backward Feature Elimination

Jeonghwan Park, Kang Li, Huiyu Zhou

Research output: Chapter in Book/Report/Conference proceedingConference contribution

307 Downloads (Pure)

Abstract

We present a new wrapper feature selection algorithm for human detection. This algorithm is a hybrid featureselection approach combining the benefits of filter and wrapper methods. It allows the selection of an optimalfeature vector that well represents the shapes of the subjects in the images. In detail, the proposed featureselection algorithm adopts the k-fold subsampling and sequential backward elimination approach, while thestandard linear support vector machine (SVM) is used as the classifier for human detection. We apply theproposed algorithm to the publicly accessible INRIA and ETH pedestrian full image datasets with the PASCALVOC evaluation criteria. Compared to other state of the arts algorithms, our feature selection based approachcan improve the detection speed of the SVM classifier by over 50% with up to 2% better detection accuracy.Our algorithm also outperforms the equivalent systems introduced in the deformable part model approach witharound 9% improvement in the detection accuracy
Original languageEnglish
Title of host publicationProceedings of the 5th International Conference on Pattern Recognition Applications and Methods
PublisherSciTePress
Pages423-430
Number of pages8
ISBN (Electronic)978-989-758-173-1
DOIs
Publication statusPublished - Feb 2016
EventInternational Conference on Pattern Recognition Applications and Methods (ICPRAM) - Rome, Italy
Duration: 24 Feb 201626 Feb 2016

Conference

ConferenceInternational Conference on Pattern Recognition Applications and Methods (ICPRAM)
CountryItaly
CityRome
Period24/02/201626/02/2016

    Fingerprint

Cite this

Park, J., Li, K., & Zhou, H. (2016). k-fold Subsampling based Sequential Backward Feature Elimination. In Proceedings of the 5th International Conference on Pattern Recognition Applications and Methods (pp. 423-430). SciTePress. https://doi.org/10.5220/0005688804230430