Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • harvard1
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
GPU cluster based brand logo detection in input video stream
KTH, School of Information and Communication Technology (ICT).
2015 (English)Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
Abstract [en]

Brand logos constitute a major part of any firm’s image and its products advertisement. It represents the firm and its quality. It also makes a positive impression in the minds of its customers. This thesis work focuses on detecting brand logos in the video sequence and to determine its effectiveness metric as perceived by the viewers. This metric could be used to compare advertisement videos which has embedded brand logos positioned in it.In this thesis, the focus is on conducting exploratory research to determine an optimal algorithm for the detection of brand logos by performing literature survey. Later the chosen algorithm is implemented on graphical processing unit and evaluation of the algorithm is carried out with respect to accuracy, precision and recall. A new approach for localization of the brand logos in video frames is developed and successfully implemented, which is capable of localizing multiple logos in video frame.Further parallelization of the algorithm is performed to accelerate the process of brand logo detection using multiple GPUs in the cluster and performance evaluation of the parallelization effort is carried out.In the process of evaluation of classifiers, the average accuracy of 0.63 with standard deviation of 0.22 is achieved in correctly classifying video frames into corresponding output class for the representative set of video test sequences. The average logo localization accuracy of 0.34 with standard deviation of 0.28, for an overlapping threshold of 90 percent is achieved with the proposed method. The fastest parallel implementation using remote CUDA for logo localization takes approximately 6.045 seconds per frame for the representative video test vector containing Fedex logo.

Place, publisher, year, edition, pages
2015. , 82 p.
Series
TRITA-ICT-EX, 2015:207
National Category
Electrical Engineering, Electronic Engineering, Information Engineering
Identifiers
URN: urn:nbn:se:kth:diva-209142OAI: oai:DiVA.org:kth-209142DiVA: diva2:1110235
Subject / course
Electronic- and Computer Systems
Educational program
Master of Science - School of Electrical Engineering (EES) - Master of Science - Research on Information and Communication Technologies
Examiners
Available from: 2017-06-15 Created: 2017-06-15 Last updated: 2017-06-15Bibliographically approved

Open Access in DiVA

No full text

By organisation
School of Information and Communication Technology (ICT)
Electrical Engineering, Electronic Engineering, Information Engineering

Search outside of DiVA

GoogleGoogle Scholar

Total: 1 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • harvard1
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf