Please use this identifier to cite or link to this item:
Scopus Web of Science® Altmetric
Type: Conference paper
Title: Hierarchical convolutional features for visual tracking
Author: Ma, C.
Huang, J.-.B.
Yang, X.
Yang, M.-.H.
Citation: Proceedings: 2015 IEEE International Conference on Computer Vision, 2015 / pp.3074-3082
Publisher: IEEE
Issue Date: 2015
ISBN: 9781467383905
ISSN: 1550-5499
Conference Name: 2015 IEEE International Conference on Computer Vision (ICCV 2015) (07 Dec 2015 - 13 Dec 2015 : Santiago, Chile)
Statement of
Chao Ma, Jia-Bin Huang, Xiaokang Yang, Ming-Hsuan Yang
Abstract: Visual object tracking is challenging as target objects often undergo significant appearance changes caused by deformation, abrupt motion, background clutter and occlusion. In this paper, we exploit features extracted from deep convolutional neural networks trained on object recognition datasets to improve tracking accuracy and robustness. The outputs of the last convolutional layers encode the semantic information of targets and such representations are robust to significant appearance variations. However, their spatial resolution is too coarse to precisely localize targets. In contrast, earlier convolutional layers provide more precise localization but are less invariant to appearance changes. We interpret the hierarchies of convolutional layers as a nonlinear counterpart of an image pyramid representation and exploit these multiple levels of abstraction for visual tracking. Specifically, we adaptively learn correlation filters on each convolutional layer to encode the target appearance. We hierarchically infer the maximum response of each layer to locate targets. Extensive experimental results on a largescale benchmark dataset show that the proposed algorithm performs favorably against state-of-the-art methods.
Keywords: Visualization; target tracking; correlation; feature extraction; semantics; spatial resolution; robustness
Rights: © 2015 IEEE
RMID: 0030076333
DOI: 10.1109/ICCV.2015.352
Appears in Collections:Computer Science publications

Files in This Item:
File Description SizeFormat 
RA_hdl_108764.pdfRestricted Access1.37 MBAdobe PDFView/Open

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.