CourtVision Augments Fan Experience with AWS Cloud Machine Learning
Clippers CourtVision is a live, augmented reality game-watching platform created by and for the Los Angeles Clippers with Second Spectrum, video tracking specialists working for the NBA.
Clippers CourtVision was developed as a way to give fans control over their game viewing experience by letting them choose different ways to augment what they see and hear with relevant information and graphics. Various machine learning, data visualisation and AR services generate the available information and graphics, which are packaged into Modes for the viewer to select.
Coach Mode creates diagrams and live layouts of plays as they happen. Player Mode displays and updates statistics above each player in real time – for example, the frame-by-frame probability of a shot going in. Mascot Mode integrates animations, graphics and FX that animate in response to conditions.
AWS is now the primary cloud computing, machine learning and artificial intelligence provider for CourtVision. New Modes and functionality that the Clippers and Second Spectrum are planning for the 2019-20 season will use AWS machine learning (ML) and data analytics services. Clippers CourtVision will also test Amazon’s SageMaker, a service for data scientists and developers to build, train and deploy ML models, to drive the statistics that will appear on live broadcasts and on-demand NBA game videos.
SageMaker is a kind of prepackaged ML lab that does most of the background work involved in ML in advance. You still need to build the models according to the information you are trying to generate, but the tools needed to prepare them for training, to connect to the training data, and to select and optimize the algorithms and framework for your application – are already in place.
To help select an appropriate algorithm, for instance, ten common ML algorithms have been pre-installed and optimised for performance. The service is pre-configured to run TensorFlow and Apache MXNet open source deep learning software frameworks as well, or you can use your own framework.
The training process for models is completely automated, including infrastructure management, scaling and tuning for accuracy. When training and tuning is finished, Amazon SageMaker deploys the model on an automatically scaled cluster of Amazon EC2 instances that are spread across availability zones to achieve high performance and high availability.
Second Spectrum has set up cameras in the 29 NBA arenas to collect 3D spatial data including ball and player locations and movements, which is stored and analysed on AWS in real time. With help from AWS’s services, the company uses that data to generate the graphical overlays on Clippers broadcasts in real time. Clippers CourtVision then uses AWS Elemental Media Services to deliver the live game-watching experience.
The CourtVision digital viewing experience is available via the FOX Sports mobile application as well as a limited Beta release, which has a choice of camera angles, live arena sounds or broadcast commentary, and a function that automatically generates facts and players’ notes. aws.amazon.com