- Zahan Shahana [email protected]
- Vinit Bodhwani [email protected]
- Niranjani Wagh [email protected]
- Prakhar Agarwal [email protected]
Highlights form an essential part of sports broadcasting due to a number of reasons like the popularity of a sport and the popularity & fan following of many players. Usually, highlights have a very long shelf life, especially for some highly competitive and nail-biting encounters. Another reason why highlights are trendy is that they package the most important and exciting elements of a match and compile them into a single video.
In soccer, there are usually many championships and leagues going on at the same time. Each match is 90 minutes long, along with many commercial advertisements and a half-time break. In a fast-paced world, it is usually difficult to catch up with each and every match. Moreover, fans of famous soccer clubs like Manchester United, Chelsea, and international teams like Brazil, Spain, and Portugal are distributed worldwide in different timezones. The best way for fans to catch up is to and remain up-to-date with their team’s matches is to watch highlights.
One can think of this as a classic use case of video summarization. In video summarization, the full-length video is converted into a shorter format such that the most critical content is preserved.
In self-generating highlights, with the help of 3D-CNN and an action recognition neural network architectures, we have created highlights for soccer matches. Our model has successfully created short video clips highlighting the three main events in a soccer match - goal, substitution, and penalty cards, with an accuracy of 95% on testing dataset.
Demo : https://drive.google.com/file/d/1lesjIldZle5C6lR5nst2rdP-3cyRt6dn/view
EDD : https://drive.google.com/file/d/1zzoGLrJcasgzW78qbmGMiOQtJUY1uYYs/view
Technical Paper : https://drive.google.com/file/d/1e6x-uPukCkps3GINtp4BUt8N92s5nP0L/view
Presentation : https://drive.google.com/file/d/1Ec_c7H-64AhMgTF4EwGw0qKVFutPcHmT/view