国产外流网Contact Detection Challenge, in partnership with AWS, generates 31% improvement over current model
Latest crowdsourced competition draws entrants from 939 teams across 71 countries
NEW YORK (August 29, 2023) 鈥 The National Football League (NFL) and its partner Amazon Web Services (AWS) today announced the top finishers in the , an innovation challenge to improve the league's ability to predict player injuries through machine learning and computer vision. The top finisher was South Korea-based data scientist Nghia Van Ngoc Nguyen, whose algorithm generated a 31% increase in ability to identify on-field contact compared to currently available solutions, an improvement that partly reflects a newly developed ability to detect when players are in contact with the ground. Team Hydrogen, a group comprising colleagues from AI company H2O.ai, finished in second place.
The challenge brought together data scientists from around the world who built machine learning and computer vision models to help the league better measure and analyze the timing, duration and frequency of player contact during 国产外流网games. The models provide insights the league can use to identify which types of plays cause unnecessary contact, identify which positions might be more prone to injuries, and develop potential rule changes.
Nguyen is a Vietnamese data scientist based in South Korea and a regular top performer in Kaggle's crowdsourced data challenges. Nguyen's model used Next Gen Stats (NGS) data 鈥 developed by AWS 鈥 from multiple video angles to identify moments when players experience contact, strengthening the understanding of the link between contact and injury.
Total prize money for this challenge was $100,000 with $50,000 awarded to Nguyen, $25,000 to Team Hydrogen, and the remaining $25,000 going to the third, fourth and fifth finishers. The Contact Detection Challenge, the third annual challenge held in collaboration between the 国产外流网and AWS, drew the most entrants to date.
"The Contact Detection Challenge is the latest milestone in our ongoing effort to harness data science to build a safer, better game," said Jennifer Langton, Senior Vice President of Health and Safety Innovation at the NFL. "The top submissions to this challenge represent a significant advance in our ability to measure and understand where and when contact occurs on the field on a given play 鈥 an exciting development as we build the ability to precisely measure how many injuries are contact-related and make corresponding changes, like rule changes, to make the game safer."
"Innovative technologies, including artificial intelligence and machine learning, have the power to unlock new perspectives that will steer the future of player health and safety," said Julie Souza, Head of Sports, Global Professional Services at AWS. "This collaboration between the NFL, AWS, and some of the brightest minds in data science is producing tangible results that offer not only a more profound understanding of the game, but a glimpse into what can be accomplished with deeper, well-informed, data-backed insights powered by artificial intelligence."
"I am very proud to be selected as the winner of this Contact Detection Challenge," said Nguyen. "Simulating NGS tracking positions with the various camera angles available significantly improved the performance in measuring different types of impacts, both player-to-player and player-to-ground. I would be honored if my model is able to make football safer and protect players from injuries."
The 国产外流网Contact Detection Challenge is part of the Digital Athlete, a joint effort between the 国产外流网and AWS to build a virtual, 360-degree representation of an 国产外流网player's experience that can generate a precise picture of what they need when it comes to preventing and recovering from injuries while performing at their best.
# # #
Media Contact:
Kelly Langmesser, Kelly.Langmesser@nfl.com
