A Comparison of Nearest Neighbor Search Algorithms for Generic Object Recognition
Lists
Only you see this
Abstract
Abstract The nearest neighbor (NN) classifier is well suited for generic object recognition. However, it requires storing the complete training data, and classifi-cation time is linear in the amount of data. There are several approaches to im-prove runtime and/or memory requirements of nearest neighbor methods: Thin-ning methods select and store only part of the training data for the classifier. Efficient query structures reduce query times. In this paper, we present an exper-imental comparison and analysis of such methods using the ETH-80 database. We evaluate the following algorithms. Thinning: condensed nearest neighbor, re-duced nearest neighbor, Baram’s algorithm, the Baram-RNN hybrid algorithm, Gabriel and GSASH thinning. Query structures: kd-tree and approximate near-est neighbor. For the first four thinning algorithms, we also present an extension to k-NN which allows tuning the trade-off between data reduction and classi-fier degradation. The experiments show that most of the above methods are well suited for generic object recognition. 1
BibTeX
@inproceedings{bajramovic2006comparison,
author = {Ferid Bajramovic and Frank Mattern and Nicholas J. Butko and Joachim Denzler},
title = {{A Comparison of Nearest Neighbor Search Algorithms for Generic Object Recognition}},
booktitle = {ACIVS},
year = {2006},
doi = {10.1007/11864349\_108},
}Loading related content...