NFL, AWS Launch New Competition to Track Helmet Impacts in Football

In 2017, researchers from Boston University studied the donated brains of 111 National Football League (NFL) players. All but one showed signs of chronic traumatic encephalopathy (CTE), a degenerative brain disease associated with behavioral problems and dementia and caused by repeated blows to the head. While CTE cannot yet be diagnosed in living players, the apparently widespread affliction of NFL players by the disease – and the ensuing backlash against the sport – has led the league to pursue creative solutions to stemming its spread. Now, the NFL has launched an artificial intelligence challenge in collaboration with Amazon Web Services (AWS), aiming to reward data scientists who help the league automatically identify players involved in a helmet impact.

The challenge, which will award prize money totaling $100,000, will provide participating data scientists with a set of video data from NFL games. The data scientists, in turn, will compete to develop the best model for identifying and tracking players automatically from the footage – and, in particular, correctly identifying players involved in a helmet impact.

“This competition is foundational in helping identify each player’s risk to injury-causing events, especially when it comes to head health,” said Jeff Miller, the NFL’s executive vice president in charge of its health and safety programs. “We are excited to enlist some of the most talented data scientists from around the world to help develop these solutions and meaningfully advance player safety.”

About seven months ago, the NFL and AWS teamed up for an antecedent challenge, which tasked participants with detecting helmet impacts using a dataset of over a thousand labeled head impact images and videos, alongside player tracking data sourced from the NFL’s Next Gen Stats (NGS) system. The current challenge is using the same dataset as the previous one, minus the mapping data provided by the NGS. The NFL is hoping that through the combination of helmet impact data and player identification data, it can establish a pipeline for identifying a given player’s helmet collision exposure across a given football play.

An example of a model identifying helmet impacts. Image courtesy of NFL/AWS.

These competitions are both part of the NFL and AWS’ efforts to develop what they call the “Digital Athlete,” a virtual representation of an NFL player aimed at predicting and preventing injuries and built on an AWS backbone.

“Using AWS’s broad range of technologies including machine learning and artificial intelligence, the NFL and AWS will generate more new insights than ever previously possible into player injuries, game rules, equipment, rehabilitation, and recovery,” said Priya Ponnapalli, senior manager of applied science for Amazon’s ML Solutions Lab. “This contest will continue that work and the data and insights collected through this project have the potential to shape the future of football.”

To learn more about the competition, visit this link.