Derinlikli nesne seziciler ve parçacik süzgeçleme ile nesne takibi

Translated title of the contribution: Object tracking by deep object detectors and particle filtering

Caner Özer, Filiz Gürkan, Bilge Günsel

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

3 Citations (Scopus)

Abstract

Developing object tracking techniques robust to blur, scale changes, occlusion and illumination changes is a challenging problem for several applications. Recently many algorithms using deep learning for visual object tracking are proposed. These algorithms mostly perform object detection without using the temporal information for video object tracking. Nevertheless, they provide high object detection accuracy as a result of an extended training scheme. However, particle filtering enables us to track the objects with a lower complexity without requiring any training, when the state transition and observation models are formulated appropriately. In this paper, tracking performance of two visual object trackers (Faster R-CNN, Mask R-CNN) and the variable rate color based particle filtering are tested on OTB-50, VOT 2016 and 2017 datasets. Benefits and deficits of both these approaches are examined. It is concluded that the deep learning methods outperform particle filtering under occlusion and scale changes, whereas particle filtering is more robust to illumination changes and blur. Integration of both approaches improves object tracking accuracy.

Translated title of the contributionObject tracking by deep object detectors and particle filtering
Original languageTurkish
Title of host publication26th IEEE Signal Processing and Communications Applications Conference, SIU 2018
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages1-4
Number of pages4
ISBN (Electronic)9781538615010
DOIs
Publication statusPublished - 5 Jul 2018
Event26th IEEE Signal Processing and Communications Applications Conference, SIU 2018 - Izmir, Turkey
Duration: 2 May 20185 May 2018

Publication series

Name26th IEEE Signal Processing and Communications Applications Conference, SIU 2018

Conference

Conference26th IEEE Signal Processing and Communications Applications Conference, SIU 2018
Country/TerritoryTurkey
CityIzmir
Period2/05/185/05/18

Bibliographical note

Publisher Copyright:
© 2018 IEEE.

Fingerprint

Dive into the research topics of 'Object tracking by deep object detectors and particle filtering'. Together they form a unique fingerprint.

Cite this