| Titre : |
AI-based image analysis for low budget volleyball teams |
| Type de document : |
Travail de fin d'études |
| Auteurs : |
Thomas VAN DER MAREN, Auteur ; Fabio SCHNEIDER, ; Marius JOLY, |
| Editeur : |
ECAM |
| Année de publication : |
2024 |
| Langues : |
Anglais (eng) |
| Mots-clés : |
Recherche & développement |
| Index. décimale : |
TFE - Ingénierie de la Santé |
| Résumé : |
The main intention of this project is to help a volleyball team and their players to improve their performance thanks to artificial intelligence tools, by making them gain time and have easier access to good analysis. An important aspect of the project is to provide a solution which is not expensive. On the market, there is a variety of different programs for analyzing sports games. The issue with those software is that they are too costly for amateur teams to acquire. Thanks to previous work, it was found that the best option to recognise the different instances like the ball, the players, etc… was to use Yolov8, which is an open-source artificial intelligence framework. A discussion with the team coach was held to evaluate the best way to tackle the project. One point that came out was that the most helpful information to know, is to know, on game replays, when which specific players play or if he/she does not, in other words, to add timestamps for players. This is important as most analysis with the players are done with replays of them. For example, if player X is only playing from minute 3 to 5, there would be no time wasted searching through the whole video, the search would be limited to the period of effective participation of the player. A large part of the work was the training of the framework, which is done on game images of the volleyball team being helped. Those images have to be labelled. This process is on good tracks but takes time as a significant number of images is used, but increasing the number of images improves precision of the model. In summary, this project is on a good way to help a low budget volleyball team and their coach by helping him analyse the video replays of games. And this, with technology that they usually can not afford, thanks to an open-source artificial intelligence framework Yolov8. Link to the video: https://youtu.be/o4WOz7jTuX4 |
AI-based image analysis for low budget volleyball teams [Travail de fin d'études] / Thomas VAN DER MAREN, Auteur ; Fabio SCHNEIDER, ; Marius JOLY, . - ECAM, 2024. Langues : Anglais ( eng)
| Mots-clés : |
Recherche & développement |
| Index. décimale : |
TFE - Ingénierie de la Santé |
| Résumé : |
The main intention of this project is to help a volleyball team and their players to improve their performance thanks to artificial intelligence tools, by making them gain time and have easier access to good analysis. An important aspect of the project is to provide a solution which is not expensive. On the market, there is a variety of different programs for analyzing sports games. The issue with those software is that they are too costly for amateur teams to acquire. Thanks to previous work, it was found that the best option to recognise the different instances like the ball, the players, etc… was to use Yolov8, which is an open-source artificial intelligence framework. A discussion with the team coach was held to evaluate the best way to tackle the project. One point that came out was that the most helpful information to know, is to know, on game replays, when which specific players play or if he/she does not, in other words, to add timestamps for players. This is important as most analysis with the players are done with replays of them. For example, if player X is only playing from minute 3 to 5, there would be no time wasted searching through the whole video, the search would be limited to the period of effective participation of the player. A large part of the work was the training of the framework, which is done on game images of the volleyball team being helped. Those images have to be labelled. This process is on good tracks but takes time as a significant number of images is used, but increasing the number of images improves precision of the model. In summary, this project is on a good way to help a low budget volleyball team and their coach by helping him analyse the video replays of games. And this, with technology that they usually can not afford, thanks to an open-source artificial intelligence framework Yolov8. Link to the video: https://youtu.be/o4WOz7jTuX4 |
|
AI-based image analysis for low budget volleyball teams
