A Context-Aware Loss Function for Action Spotting in Soccer Videos

TL;DR

This paper introduces a novel context-aware loss function for action spotting in soccer videos, improving temporal localization accuracy by leveraging surrounding context, and demonstrates its effectiveness across multiple datasets and applications.

Contribution

The paper proposes a new loss function that incorporates temporal context for better action spotting, with extensive benchmarking and analysis on soccer and activity datasets.

Findings

12.8% improvement over baseline on SoccerNet

Effective generalization to activity detection in ActivityNet

Enhanced temporal understanding for automatic highlights

Abstract

In video understanding, action spotting consists in temporally localizing human-induced events annotated with single timestamps. In this paper, we propose a novel loss function that specifically considers the temporal context naturally present around each action, rather than focusing on the single annotated frame to spot. We benchmark our loss on a large dataset of soccer videos, SoccerNet, and achieve an improvement of 12.8% over the baseline. We show the generalization capability of our loss for generic activity proposals and detection on ActivityNet, by spotting the beginning and the end of each activity. Furthermore, we provide an extended ablation study and display challenging cases for action spotting in soccer videos. Finally, we qualitatively illustrate how our loss induces a precise temporal understanding of actions and show how such semantic knowledge can be used for automatic highlights generation.