Deep Learning for Video Analytics in Sport Events

Type Start End
National Feb 2018 Jan 2021
Responsible URL
Ferran Marqués Doctorats Industrials de la Generalitat de Catalunya

Reference

Reference: 2017 DI 064

This Phd thesis will exploit the latest advances in artificial intelligence to the domain of video analysis of sport events. Mediapro is a major worldwide producer of sport event for broadcasting and aims at enriching the user experience with new and richer statistics of sports events, together with automatic production and indexing of video content. The Image Processing Group from the Universitat Politecnica de Catalunya (UPC) is a pioneering center in the latest wave of artificial intelligence powered with deep learning. Since 2013, it has been exploring the applications of deep neural networks for computer vision, with major publications and innovative courses both in class and online.

The research project will aim at covering the requirements of Mediapro based on convolutional and recurrent neural networks capable of analising video streams and fuse them with additional data sources, such as wearable sensors on the football players, accelerometers on the motorcycles or audience feedback through social media.

Publications

Girbau A. Sports broadcasting and multiple object tracking with deep learning methods Rius I, Giró-i-Nieto X, Marqués F. Signal Theory and Communications. 2021 ;PhD.
Ventura C, Bellver M, Girbau A, Salvador A, Marqués F, Giró-i-Nieto X. RVOS: End-to-End Recurrent Network for Video Object Segmentation. In: CVPR. CVPR. Long Beach, CA, USA: OpenCVF / IEEE; 2019. (5.76 MB)

Collaborators

Andreu Girbau PhD Candidate andreu.girbau@upc.edu
Xavier Giró Associate Professor xavier.giro@upc.edu
Ferran Marqués Professor ferran.marques@upc.edu