| 講演抄録／キーワード |
| 講演名 |
2024-03-21 14:25
Bunch of Tricks for Improving Shuttlecock Detection from Badminton Videos
○Muhammad Abdul Haq（TMU）・Shuhei Tarashima（NTT Com）・Norio Tagawa（TMU） |
| 抄録 |
（和） |
(まだ登録されていません) |
| （英） |
Accurate identification of the shuttlecock is necessary for video analysis in badminton matches, but, it remains difficult to determine the small and rapidly moving shuttlecock from videos. In this study, we present multiple enhancements that can be integrated into the advanced shuttlecock identification algorithm, TrackNetV2. Our suggested techniques involve implementing residual networks, leaky ReLU activation functions, and post-addition batch normalization. These techniques are all derived from deep learning theory. The contribution of this work is to apply the well-known techniques to shuttlecock detection successfully, and this is the first attempt in shuttlecock detection literature. The results of our studies on a publicly available badminton dataset show that our proposed techniques significantly enhance the performance of TrackNetV2 in terms of precision, recall, F1 score, and accuracy. Specifically, the precision increases from 84.8% to 88.2%, recall improves from 81.8% to 83.1%, F1 score improves from 83.3% to 85.3%, and accuracy improves from 75.8% to 78.1%. |
| キーワード |
（和） |
/ / / / / / / |
| （英） |
Shuttlecock detection / Deep learning / Training tricks / / / / / |
| 文献情報 |
映情学技報, vol. 48, no. 14, SIP2024-4, pp. 8-11, 2024年3月. |
| 資料番号 |
SIP2024-4 |
| 発行日 |
2024-03-14 (SIP) |
| ISSN |
Online edition: ISSN 2424-1970 |
| PDFダウンロード |
|
|