Video Summarization for Multiple Sports Using Deep Learning

This paper proposes a computationally inexpensive method for automatic key-event extraction and subsequent summarization of sports videos using scoreboard detection. A database consisting 1300 images was used to train supervised-learning based object detection algorithm, YOLO (You Only Look Once). Then, each frame the video, once detected YOLO, cropped out image. After this, image processing techniques were applied on reduce noise false positives. Finally, processed passed through an OCR (Optical Character Recognizer) get score. rule-based algorithm run output generate timestamps key-events game. The proposed is best suited people who want analyse games precise occurrence important events. performance design tested Bundesliga, English Premier League, ICC WC 2019, IPL Pro Kabaddi League. An average F1 Score 0.979 achieved during simulations. trained five different classes three separate (Soccer, Cricket, Kabaddi). implemented python 3.7.