Camera Calibration and Player Localization in SoccerNet-v2 and Investigation of their Representations for Action Spotting

TL;DR

This paper introduces a large-scale soccer broadcast dataset, distills a commercial calibration tool into a neural network, and demonstrates improved action spotting performance in SoccerNet-v2 by leveraging new camera calibration and player localization representations.

Contribution

It presents the first large-scale calibration dataset for soccer videos, a distilled neural network for calibration, and integrates these into action spotting to achieve state-of-the-art results.

Findings

Achieved new state-of-the-art in soccer action spotting.

Provided three novel representations for camera calibration and player localization.

Released a publicly available calibration network trained on SoccerNet.

Abstract

Soccer broadcast video understanding has been drawing a lot of attention in recent years within data scientists and industrial companies. This is mainly due to the lucrative potential unlocked by effective deep learning techniques developed in the field of computer vision. In this work, we focus on the topic of camera calibration and on its current limitations for the scientific community. More precisely, we tackle the absence of a large-scale calibration dataset and of a public calibration network trained on such a dataset. Specifically, we distill a powerful commercial calibration tool in a recent neural network architecture on the large-scale SoccerNet dataset, composed of untrimmed broadcast videos of 500 soccer games. We further release our distilled network, and leverage it to provide 3 ways of representing the calibration results along with player localization. Finally, we exploit those representations within the current best architecture for the action spotting task of SoccerNet-v2, and achieve new state-of-the-art performances.