Next Generation Sport Systems - AI-Based Video Analysis, Processing, and Delivery
Interested in sport technology for videos used by elite leagues and live broadcasters? We have for long researched sports technology, both for distributing the video of live broadcasts and managing the archives in Norway and Sweden. We develop systems that are actively used and tested on real users, where the users can see highlights and summaries, get recommendations, collect favorite events, make personalized playlists, etc.
Interested in sport technology for videos used by elite leagues and live broadcasters? We have for long researched sports technology, both for distributing the video of live broadcasts and managing the archives in Norway and Sweden. We develop systems that are actively used and tested on real users, where the users can see highlights and summaries, get recommendations, collect favorite events, make personalized playlists, etc.
Goal
There are a lot of open questions and issues to address like how to make:
- Automated AI-based event detection systems: See for example: home.simula.no/~paalh/publications/files/ism2020-olav.pdf , doi.org/10.3390/make3040051. Find events like goals, cards, tackles, dribbles automatically in the video stream. This is usually done in a tedious manual operation, but can video/audio content analysis help to automate this operation? Can off-screen events also be detected? How about certain types of unsportsmanlike conduct, which might under normal circumstances go unnoticed by the referee?
- Automated AI-based clipping systems: See for example : www.mdpi.com/2504-4990/3/4/49 , ieeexplore.ieee.org/document/9666063. It is important to have good quality clips. At the same time, a lot of time goes into clipping events, and it is important to publish events fast. Here, we aim to find these clipping points automatically and provide an automated service by automatically finding positions in the datastream where there is a camera change, a logo transition, a break in the play, etc.
- Automated AI-based thumbnail selection systems: See for example: dl.acm.org/doi/10.1145/3524273.3528182. Thumbnails are the first view / teaser of an event with the puropse of attracting attention and interest of an event. Fainting a good one is timeconsuming, and here you should automate this challenge.
- Automated AI-based summarization systems & story-telling: See for example: dl.acm.org/doi/10.1145/3552463.3557019. Making summaries and highlights for various purposes, like the sport-news, takes a lot of time. Can we automate this process? Given that we have a lot of events, how can they be compiled into a summary-higlight for the TV2 or NRK news?
- Sports video captioning: Automatic generation of timestamped annotations, going beyond the task of event detection. Generalized methodology for creating scene descriptions. The goal of this task is to help curate a large dataset of soccer game highlights which will allow for the automatic creation of game summaries using AI. In particular, our focus is on scraping websites such as www.bbc.com/sport/football, www.flashscore.com/football/, www.goal.com/, www.vg.no/sport/fotball/, www.aftenposten.no/sport/fotball/, www.dagbladet.no/emne/fotball for game commentary and captions (list of timestamped entries such as: www.flashscore.com/match/6R6OiSj5/#/match-summary/live-commentary/0). We would like to include as many leagues and as many seasons as possible. The new dataset will possibly be used for a new challenge under the SoccerNet umbrella and curators will be credited.
- Investigate solutions for attracting users: How can we attract users to such video systems? How to boost activity? Can we gamify the user interactions? Make reward systems? Clever ways of sharing on social media?
We collaborate with the leagues and elite clubs in Norway and Sweden, the football associations, Telenor, broadcasters, Forzasys (a streaming technology company), Telenor and several international partners.
Want to be part of our research group and improve the current systems? Contact us, and let’s see if we can find a thesis topic that matches your interests!
Learning outcome
- Video analysis
- Machine learning / AI
- User studies
- Building systems
Qualifications
- Hard working, motivated
- Interested in learning (the rest can be learned during the thesis work)
Supervisors
- Pål Halvorsen
- Michael Riegler
- Cise Midoglu
Collaboration partners
- ForzaSys AS