Digitala Vetenskapliga Arkivet

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
Football shot detection using convolutional neural networks
KTH, School of Electrical Engineering and Computer Science (EECS).
2024 (English)Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesisAlternative title
Detektion av skott i fotboll med hjälp av convolutional neural networks (Swedish)
Abstract [en]

The application of computer vision in football is becoming more and more attractive. Computer vision can track players and actions allowing statistics and analysis of the sport. This study aims to recognize shots in football video data using computer vision. Recognizing actions in video data has been explored using various techniques. Convolutional Neural Networks, CNNs, have dominated computer vision by proving great at understanding image data. However, extending CNNs to understand video data is a difficult task. In recent years 3D CNNs have been developed to understand the temporal information in video data, showing promising results in the field of action recognition. In this study, 3D CNNs are compared to standard CNNs in their ability to predict shots in football video data. Football Analytics Sweden AB collects and owns the data which consist of short clips of shots and non-shots. 3D CNNs were found to outperform 2D CNNs in recognizing shots, achieving an accuracy of 85% on the test dataset. The results show that 3D CNNs have the potential to distinguish between different actions in football video data and could be used in further analysis of football matches using computer vision.

Abstract [sv]

Användning av maskininlärning och datorseende inom fotboll blir alltmer attraktiv. Datorseende möjliggör detektion av spelare och händelser, vilket kan användas för analys och statistik inom sporten. Denna studie avser detektion av skott i fotbollsvideodata med hjälp av datorseende. Att känna igen handlingar i videodata har undersökts med olika tekniker. Convolutional Neural Networks, CNNs, har dominerat datorseende genom att vara mycket effektiva på att förstå bilddata. Att utöka CNNs för att förstå videodata är dock en svår uppgift. Under de senaste åren har 3D-CNN utvecklats för att förstå hur datan ändras över tid i videodata, och har visat lovande resultat inom detektering av handlingar. I denna studie jämförs 3D-CNN med vanliga CNN i deras förmåga att detektera skott i fotbollsvideodata. Football Analytics Sweden AB samlar in och äger data som består av korta klipp av skott och icke-skott. 3D-CNN visade sig överträffa 2D-CNN i att känna igen skott och uppnådde en noggrannhet på 85% på testdatan. Resultaten visar att 3D-CNN har potential att skilja mellan olika handlingar i fotbollsvideodata och kan användas i vidare analys av fotbollsmatcher med hjälp av datorseende.

Place, publisher, year, edition, pages
2024. , p. 41
Series
TRITA-EECS-EX ; 2024:827
Keywords [en]
Machine Learning, Computer Vision, Action Detection
Keywords [sv]
Maskininlärning, Datorseende, Handlingsdetektion
National Category
Computer and Information Sciences
Identifiers
URN: urn:nbn:se:kth:diva-360335OAI: oai:DiVA.org:kth-360335DiVA, id: diva2:1940034
External cooperation
Football Analytics Sweden AB
Supervisors
Examiners
Available from: 2025-02-27 Created: 2025-02-25 Last updated: 2025-02-27Bibliographically approved

Open Access in DiVA

fulltext(2931 kB)90 downloads
File information
File name FULLTEXT02.pdfFile size 2931 kBChecksum SHA-512
f644918e3af53f5e4f35c0169566f510584123c31841f93eeb9073f6ca4f3de3f1744ad1e8b39e24137ffd77e0b7bff690de0095c3ea4de0f5ef629be3389972
Type fulltextMimetype application/pdf

By organisation
School of Electrical Engineering and Computer Science (EECS)
Computer and Information Sciences

Search outside of DiVA

GoogleGoogle Scholar
Total: 90 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: 569 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