IIIT Hyderabad Publications |
|||||||||
|
A Novel framework for Fine Grained ActionRecognition in SoccerAuthors: Ganesh Yaparla,Sriteja Allaparthi,Sai Krishna Munnangi,Garimella Ramamurthy Conference: 15th International Work-Conference on Artificial Neural Networks Location Gran Canaria, Spain. Date: 2019-06-12 Report no: IIIT/TR/2019/17 AbstractSports analytics have become a topic of interest in the field ofArtificial intelligence. With the availability of huge volumes of high leveldata, significant progress has been made in the domain of action recog-nition in the past. Though video based action recognition has progressedwell using state of the art deep learning techniques, its applications arelimited to some higher level actions like throwing, jumping, running etc.There has been some work in fine-grained action recognition technique,such as, identification of type of throws in Basketball, and the type of aplayer’s shots in Tennis. However with larger play field and with manyplayers on field, multi player sports such as Soccer, Rugby, Hockey andetc. pose bigger challenges and remain unexplored. These games in gen-eral are live fed through field view cameras or skycams which aren’tstationary. For these reasons, we chose to recognize player’s actions inthe game of Soccer and thereby, explore the capabilities of existing ar-chitectures and deep neural networks for these kind of games. Our maincontributions are the proposed framework that can automatically recog-nize actions of players in live football game which will be helpful for textquery based video search, for extracting stats in a football game and togenerate textual commentary and the Soccer-8k dataset which consistsof different action clips in the soccer play Full paper: pdf Centre for Communications |
||||||||
Copyright © 2009 - IIIT Hyderabad. All Rights Reserved. |