Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • 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
An inquiry into the efficacy ofconvolutional neural networks in low-resolution video feeds for object detection
KTH, School of Engineering Sciences (SCI), Mathematics (Dept.), Mathematical Statistics.
2019 (English)Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesisAlternative title
En undersökning gällande effektiviteten av convolutional neurala nätverk i låg-kvalitets video-strömmar för objekt detektion (Swedish)
Abstract [en]

In this thesis, various famous models have been investigated and compared to a custom model for people detection in low resolution video feeds. YOLOv3 and SSD in particular are famous models which have, at their time, produced state of the art results on competitions such as ImageNet and COCO. The performance of all models have been compared on speed and accuracy where it was found that YOLOv3 was the slowest and SSD was the fastest. The proposed model was superior in accuracy to both of the aforementioned architectures which can be attributed to addition of newer techniques from research such as leaving activations out and having a carefully balanced loss function. The results seem to suggest that the proposed model is implementable for real-time inference using cheap hardware such as a raspberry pi 3B+ coupled with one or more AI accelerator stickssuch as the Intel Neural Compute Stick 2 and that the networks are usable for detection even in bad video streams.

Abstract [sv]

I denna uppsats så har olika kända modeller undersökts och jämförts med en ny modell för människodetektering i lågkvalitets videoströmmar. YOLOv3 och SSD mer specifikt är kända modeller som, för sin tid, producerade topp resultat på tävlingar såsom ImageNet och COCO. Prestandan för alla modeller jämfördes medavseende på hastighet och träffsäkerhet där det hittades att YOLOv3 var den långsammaste och SSD var den snabbaste. Den förslagna modellen var träffsäkrare än båda tidigarenämnda modeller vilket kan attribueras till att nya tekniker från forskning har tillämpats såsom att låta vissa aktiveringsfunktioner utebli och att ha en försiktigt balanserad förlust funktion. Resultaten pekar mot att den förslagna modellen kan implementeras för bruk i real tid på billig hårdvara såsom en Raspberry pi 3B+ tillsammans med en eller flera AI accelerations stickor så som Intel Neural Compute Stick 2 samt att nätverken är användbara för detektion även i dåliga videoströmmar.

Place, publisher, year, edition, pages
2019.
Series
TRITA-SCI-GRU ; 2019:097
Keywords [en]
Neural networks, statistics, applied mathematics, person detection, CNN
Keywords [sv]
Neurala nätverk, statistik, tillämpad matematik, person detektion, CNN
National Category
Probability Theory and Statistics
Identifiers
URN: urn:nbn:se:kth:diva-252704OAI: oai:DiVA.org:kth-252704DiVA, id: diva2:1320415
External cooperation
Stanley Security Sverige
Subject / course
Mathematical Statistics
Educational program
Master of Science - Applied and Computational Mathematics
Supervisors
Examiners
Available from: 2019-06-04 Created: 2019-06-04 Last updated: 2019-06-04Bibliographically approved

Open Access in DiVA

fulltext(2638 kB)55 downloads
File information
File name FULLTEXT01.pdfFile size 2638 kBChecksum SHA-512
76b6f2d8a0ba1bed3a1a97e61c03ffccf38b7e42c9e491f3b36af63ecbd6dee14b6b08d6f70a1d47a8971c8c6193f63a09f608c2848599726e7d49bbc5b19934
Type fulltextMimetype application/pdf

By organisation
Mathematical Statistics
Probability Theory and Statistics

Search outside of DiVA

GoogleGoogle Scholar
Total: 55 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

urn-nbn

Altmetric score

urn-nbn
Total: 98 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • 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