Sports produce a slew of data. In a game of cricket, for example, each play generates millions of video-frame data points for a sports analyst to scrutinize. Automated video processing is revolutionary in sports, and we are excited to build more advanced models and pipelines to keep the revolution going. — Masoumeh Izadi, MD of deep-tech startup TVConal
Remember the city-scale NVIDIA Metropolis video analytics platform that helps drive a smart city vision by delivering valuable insights to improve traffic, public safety and build an AI City of the future? In essence, a vision-based processing platform, it can be scaled for any other image analysis-based needs such as sports analytics. NVIDIA Metropolis has been immensely helpful for sports coaching assistance through video analytics processing to derive statistics, player form and more through the likes of prior startups such as Pixellot and Track160.
A Singapore-based company TVConal (which stands for Television Content Analytics) and a member of NVIDIA’s Inception program that support startups with cutting-edge technology has also joined in the fray to use NVIDIA’s AI and computer vision processing to power its video analytics platform with a focus on popular sports in Asia, such as badminton, cricket, football and tennis.
Its platform – powered by the NVIDIA Metropolis application framework for vision AI – can detect important in-game events, model athlete behaviour, make movement predictions and more. It helps dissect the minute details in sports, enabling teams to make smarter decisions on the field and even inform umpires of illegal player actions that might have gone unnoticed by the human eye and empower them to take split-second decisions with accurately analysed video data that could change the tide of the game.
NVIDIA GPU-accelerated compute resources used in TVConal’s platform include the Jetson platform for AI at the edge, RTX 3090 workstations on-prem and Tesla V100 and A100 in the cloud. Utilising NVIDIA DeepStream SDK to simplify video processing pipelines, NVIDIA pre-trained models and TAO toolkit to accelerate AI training and the NVIDIA TensorRT SDK to optimize inferencing, these allow the TVConal team to process live video and audio streams in real-time, which is the necessary speed to match video frame rates. In addition, the TensorRT library helped TVConal convert its machine learning models to more quickly process data, while maintaining accuracy.
Thanks to the use of the full breadth of NVIDIA tools (and hardware), including pre-trained AI models, this allows TVConal to compress training custom AI models from weeks to days.
The advanced hardware and software platform is no doubt is one of the highlights that allows TVConal to deliver a compelling near real-time sports analysis solution that is critical to enabling smarter decisions that guide sports teams, leagues and even broadcasters to deliver vital performance insights.
