Tennis event detection struggles beyond broadcast camera views
Can Event Detectors Leave the Broadcast Camera? Evidence from Tennis
Computer Vision and Pattern RecognitionArtificial Intelligence
Summary
Detecting events like ball bounces and racket hits in tennis videos is easy when using broadcast footage, but much harder with videos from other angles like phones or court-level cameras. The authors created a benchmark with annotations from different viewpoints and tested several detection methods. They found that models trained on broadcast videos perform poorly on other views, while a new method, RallyArb, works better on some camera angles but has timing issues with hits. This work highlights the challenges of detecting tennis events outside traditional broadcast views.
What this means in practice
- •For sports analytics teams: Improve event detection models to work reliably on mobile and court-level camera videos during tennis matches.
- •For video technology developers: Develop tennis replay and highlight tools that accurately detect bounces and hits from diverse camera angles beyond broadcast footage.$Commercial implications: This paper enables new video analysis products for tennis broadcasters and content creators by improving event detection across various viewpoints.
Authors
David Soeiro-Vuong, Andre-Louis Rochet
Abstract
Ball bounces and racket hits are the events from which tennis statistics are derived, and on monocular video the two are easy to confuse. We release a benchmark of 345 bounce and hit annotations at frame level on eight clips that mix broadcast, fixed elevated, phone and court-level viewpoints, with a frozen development and test split, a confusion-aware matcher, clip-level bootstrap intervals and a blind second annotation of three clips. On it we score supervised precise spotters with published broadcast weights, two zero-shot vision-language models, and RallyArb, an arbitration layer over a ball track and player poses that fits no parameter on event labels. The supervised spotters score near-perfectly on broadcast footage and almost nothing on phone and court-level footage, where the pooled bounce F1 of E2E-Spot is 0.130. RallyArb keeps a bounce F1 above 0.8 on the three phone and court-level clips, two of which were used to tune it, and finds 8 of 11 bounces on the held-out one, but it places hits too far from the contact frame to match human timing. The vision-language models detect events on every viewpoint and miss the contact frame by about 0.1 s. Fine-tuned leave-one-clip-out from its broadcast weights, the spotter reaches about 0.8 bounce F1 on two phone and court-level clips and stays at 0.2 on a grazing phone angle. The annotations, the evaluator and the frozen predictions of every scored row are released.