Current Result Document :
| 한글제목(Korean Title) |
농구 중계 영상에서 득점 검출 시스템 |
| 영문제목(English Title) |
Goal Detection System in Basketball Video |
| 저자(Author) |
서남진
이현섭
최희열
Namjin Seo
Hyunsub Lee
Heeyoul Choi
|
| 원문수록처(Citation) |
VOL 48 NO. 02 PP. 0653 ~ 0655 (2021. 12) |
한글내용
(Korean Abstract) |
|
영문내용
(English Abstract) |
In the basketball field, many statistics are used to evaluate the performance of players and teams. Among the statistics, goal and shoot are some of the fundamental statistics in the game. In this paper, we propose an autoregressive model to classify a goal at each frame in a basketball game video. We implement an CNN-based classifier that classifies three states (e.g., goal, shoot, and no event) from the hoop area detected by object detection frameworks. However, dealing with high dimensionalities of video data is not easy. One of the challenges is aggregating information over time steps. Events in video occur over variable frames, and the classifier should refer to the previous information. Another challenge is class imbalance. The goal in a game is much less than other classes such as "no event" or "shoot". In training model, we introduce an effective model to aggregate the information over time steps and overcome the class imbalance between the goal and the other classes. As the results, we show our approach can achieve 0.939 F1 score on our custom dataset generated by object detection frameworks and outperform the baseline model. |
| 키워드(Keyword) |
|
| 파일첨부 |
PDF 다운로드
|
