This article aims to present the benefits of using a deep learning approach to perform race analyses during domestic and international swimming championships. The procedure currently used to perform these race analyses is mostly manual and requires important human resources to annotate the videos and produce the performance reports. Recent technological and scientific developments now allow using ultra high quality cameras (4K) and machine learning algorithms to automatise the detection of the key events and greatly improve the video processing. Such a process helps the collection of data with a higher accuracy, the deployment of a more flexible and reliable setup, the access to the more variables such as the swimmers‘ instantaneous position and velocity, and the redistribution of the human resources to more effective actions.
"A NEW PARADIGM TO DO AND UNDERSTAND THE RACE ANALYSES IN SWIMMING: THE APPLICATION OF CONVOLUTIONAL NEURAL NETWORKS,"
ISBS Proceedings Archive: Vol. 37
, Article 112.
Available at: https://commons.nmu.edu/isbs/vol37/iss1/112