MOT, as well as the majority of Computer Vision tasks, is a resource-intensive activity. For our solution we implemented FairMOT 10, which performs the two tasks of player detection and re-identification in a single network to significantly improve inference speed. There are a wide range of open source MOT solutions offered, such as Deep SORT 7, DEFT 8 and GSDT 9. This problem of object detection and re-identification is at the core of Multiple Object Tracking (MOT) in computer vision. To extract value from video footage of a sports match, not only detecting all player positions on the pitch is needed but also their corresponding tracks over time. In the following section we will focus on the technical challenges that are faced in applying computer vision to field hockey. Our previous blog on the AI4Animals solution 6 touched upon a general overview of aspects that are needed for a successful implementation of computer vision solutions that also apply for the solution discussed here. Currently, multiple use cases are ready to be rolled out that can disrupt the sponsoring market as we know it, both in new ways of sponsoring as in measuring effectiveness. Lastly, to increase sponsorship value HockEye introduces new ways of sponsoring when using real-time, API-driven predictions. We believe this has the potential to revolutionize the game of field hockey as we know it in a broader sense. By making the solution API driven, the number of fan digital outlets, where developers can build upon, is huge. ![]() Next to improving team performance, HockEye forms the base to increase fan engagement. ![]() Extracting positional data from hundreds of hours of video data augments and greatly enhances current best practices in field hockey. Game strategies and tactical decisions are based on the interaction between teams and the positions of the players on the field.Information from GPS wearables is only available from one team, which leaves a blind spot for the opposing team.The current process to extract game information through manual tagging is time consuming and thus expensive, which can be alleviated by automatic tracking and event prediction.Starting with team performance, our HockEye solution detects and tracks hockey players in order to tackle several challenges faced in current high performance field hockey, for example: There are 137 National Associations of hockey spread over all five continents and TV broadcasts of the 2018 World Cup were available in over 194 countries 5. With field hockey having such a solid fanbase in many countries around the world, and having an innovation name to uphold, it is in a great position to deploy AI to revolutionize the sport by maximizing the interaction between team performance, fans and sponsors. ![]() Not only to increase the performance, but also to better engage fans with live stats and on-demand broadcasting, and thereby increase sponsorship value. These sports have invested heavily in the use of AI (see our latest blog ‘Sports goes web-scale: New insights, new fans’ 4). On the AI technology side, however, field hockey has some catching up to do compared to high-budget sports like American football, baseball, basketball, golf and football. This is illustrated by frequent rule changes to increase attractiveness and safety of the game, advancements in equipment and early adoption of video referees. Field hockey is widely known to be an innovative sport compared to other (bigger) sports like football.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |