swimming

πŸŠβ€β™€οΈ Swimming Research

Contact: romain.vuillemot@ec-lyon.fr

What do we do?

Swimming research at Centrale Lyon on the analysis of videos of elite swimming races in competition. We developed and relseased serveral tools that were built for the 2024 Paris Olympics and funded by ANR NePTUNE and CNRS.

See also the πŸ“ Table Tennis Research that was conducted during the same period at Centrale Lyon.

πŸ“Š Data

SwimTrack Dataset which is a dataset of annotated videos of elite swimming races

SportsVideo a follow-up of SwimTrack Dataset with more videos and annotations (swimming and table tennis).

βš™οΈ Code

Aquanalysis: Analyze annotations from Aquanote.

Aquanote is an open-source annotation tool for race swimming videos (using one or multiple static cameras).

ntt is a set of wrappers around video and image processing libraries (OpenCV, FFmpeg, etc.) to facilitate video analysis and preparation for annotation tasks.

pipeoptz is a set of tools to optimize video processing pipelines (for object detection, tracking, etc.).

swimmers_detection for swimmers position detection in swimming race videos.

swimmers_cycles_counting for counting the number of cycles (strokes) performed by swimmers in swimming race videos.

sports_field_registration for registering sports fields (swimming pool, football field, etc.) in videos.

πŸš€ Tools

The main tool we developed is Aquanote an open-source annotation tool for race swimming videos (using one or multiple static cameras). The tool allows to annotate the position of swimmers, their strokes, and other events (turns, finishes, etc.) in the videos. It also allows to visualize the annotations and to export them in various formats for further analysis.

Aquanote Screenshot

Publications

Nicolas Jacquelin (2022). Analyse Automatique de VidΓ©os de Course de Natation Elite. PhD Thesis 2022. https://hal.science/hal-03936053/file/paper6876.pdf

Nicolas Jacquelin, Theo Jaunet, Romain Vuillemot, Stefan Duffner. SwimTrack: Swimmers and Stroke Rate Detection in Elite Race Videos. MediaEval 2023. https://hal.science/hal-03936053/file/paper6876.pdf

People

Team:

Press

Acknowledgments