Publication Details
Real-time Algorithms of Object Detection using Classifiers
Hradiš Michal, Ing., Ph.D. (DCGM FIT BUT)
Zemčík Pavel, prof. Dr. Ing. (DCGM FIT BUT)
WaldBoost, Object Detection, Classification Cost, EnMS, Neighborhood Suppression, Acceleration, GPU, SIMD, FPGA
Real-time object detection is currently expanding field of computer vision. One of most popular methods for the object detection is based on exploitation of statistical classifiers (namely those based on AdaBoost algorithm). This contribution presents methods for acceleration of object detection based on the AdaBoost. First it describes image pre-processing and the learning of classifiers. Finally, the contribution presents algorithmic accelerations of the detection process and effective implementations of classification on various architectures - CPU/SSE, GPGPU and FPGA. Accelerated detectors achieve high performance compared to state of the art solutions and are suitable for real-time applications.
@INBOOK{FITPUB9883,
author = "Roman Jur\'{a}nek and Michal Hradi\v{s} and Pavel Zem\v{c}\'{i}k",
title = "Real-time Algorithms of Object Detection using Classifiers",
pages = "1--22",
booktitle = "Real-Time System",
year = 2012,
location = "Rijeka, HR",
publisher = "InTech - Open Access Publisher",
ISBN = "9789535105107",
language = "english",
url = "https://www.fit.vut.cz/research/publication/9883"
}