Skip navigation
Skip navigation

Hyperparameter Optimization for Tracking with Continuous Deep Q-Learning

Dong, Xingping; Shen, Jianbing; Wang, Wenguan; Yu, Liu; Shao, Ling; Porikli, Fatih

Description

Hyperparameters are numerical presets whose values are assigned prior to the commencement of the learning process. Selecting appropriate hyperparameters is critical for the accuracy of tracking algorithms, yet it is difficult to determine their optimal values, in particular, adaptive ones for each specific video sequence. Most hyperparameter optimization algorithms depend on searching a generic range and they are imposed blindly on all sequences. Here, we propose a novel hyperparameter...[Show more]

dc.contributor.authorDong, Xingping
dc.contributor.authorShen, Jianbing
dc.contributor.authorWang, Wenguan
dc.contributor.authorYu, Liu
dc.contributor.authorShao, Ling
dc.contributor.authorPorikli, Fatih
dc.coverage.spatialSalt Lake City, United States
dc.date.accessioned2020-09-16T03:20:57Z
dc.date.createdJune 18-22 2018
dc.identifier.isbn978-153866420-9
dc.identifier.urihttp://hdl.handle.net/1885/210512
dc.description.abstractHyperparameters are numerical presets whose values are assigned prior to the commencement of the learning process. Selecting appropriate hyperparameters is critical for the accuracy of tracking algorithms, yet it is difficult to determine their optimal values, in particular, adaptive ones for each specific video sequence. Most hyperparameter optimization algorithms depend on searching a generic range and they are imposed blindly on all sequences. Here, we propose a novel hyperparameter optimization method that can find optimal hyperparameters for a given sequence using an action-prediction network leveraged on Continuous Deep Q-Learning. Since the common state-spaces for object tracking tasks are significantly more complex than the ones in traditional control problems, existing Continuous Deep Q-Learning algorithms cannot be directly applied. To overcome this challenge, we introduce an efficient heuristic to accelerate the convergence behavior. We evaluate our method on several tracking benchmarks and demonstrate its superior performance1.
dc.format.mimetypeapplication/pdf
dc.language.isoen_AU
dc.publisherIEEE
dc.relation.ispartof31st Meeting of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2018
dc.rights© 2018 IEEE
dc.sourceProceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition
dc.titleHyperparameter Optimization for Tracking with Continuous Deep Q-Learning
dc.typeConference paper
local.description.notesImported from ARIES
local.description.refereedYes
dc.date.issued2018
local.identifier.absfor080104 - Computer Vision
local.identifier.ariespublicationu3102795xPUB1509
local.publisher.urlhttps://www.ieee.org/
local.type.statusPublished Version
local.contributor.affiliationDong, Xingping, Beijing Institute of Technology
local.contributor.affiliationShen, Jianbing, Beijing Lab of Intelligent Information Technology
local.contributor.affiliationWang, Wenguan, Beijing Institute of Technology
local.contributor.affiliationYu, Liu, Beijing Institute of Technology
local.contributor.affiliationShao, Ling, University of East Anglia
local.contributor.affiliationPorikli, Fatih, College of Engineering and Computer Science, ANU
local.description.embargo2037-12-31
local.bibliographicCitation.startpage518
local.bibliographicCitation.lastpage527
local.identifier.doi10.1109/CVPR.2018.00061
local.identifier.absseo899999 - Information and Communication Services not elsewhere classified
dc.date.updated2020-06-23T00:53:54Z
local.identifier.scopusID2-s2.0-85061742623
CollectionsANU Research Publications

Download

File Description SizeFormat Image
01_Dong_Hyperparameter_Optimization_2018.pdf948.25 kBAdobe PDFThumbnail


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

Updated:  19 May 2020/ Responsible Officer:  University Librarian/ Page Contact:  Library Systems & Web Coordinator