Name: Marina Ivasic-Kos
Affiliation: Faculty of Informatics and Digital Technologies, University of Rijeka
Primary research interest: Computer Vision, Machine and Deep Learning, Object Detection, Object Tracking, Action Recognition
Title of the lecture: Content analysis of sports scenes using deep neural networks
Keywords: Object Detection, Object Tracking, Action Recognition, Deep Neural Networks, Handball
Summary: This lecture deals with the application of deep neural networks in the analysis of sports scenes such for object detection, multi-object tracking and action recognition and localization. The aim of the analysis of sports scenes is multiple, from helping the player and coach in monitoring the performance and progress of players, analysis of techniques, assistance in refereeing and the like.
The lecture will provide basic information about neural networks and convolutional deep neural networks and present several experiments dealing with analysis of handball scenes using state-of-the-art deep neural networks, that include player and ball detection, player tracking, and activity recognition and localization in video. Handball scenes are very demanding for object detectors and trackers, since players move quickly on the field under artificial lighting with a crowded background, change the position and distance from the camera, and often overlap so can very well present the advantages and limits of current deep learning models.
In conclusion, open issues and challenges in applying deep learning methods, in such a dynamic sports environment, will be discussed.