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, according to Masoumeh Izadi, managing director of deep-tech startup TVConal.
The Singapore-based company uses NVIDIA AI and computer vision to power its sports video analytics platform, which enables users — including sports teams, leagues and TV broadcasters — to gain performance insights from these massive amounts of data in real time.
Short for Television Content Analytics, TVConal provides video analytics for a variety of sports, with a focus on cricket, tennis, badminton and football.
Its platform — powered by the NVIDIA Metropolis application framework for vision AI — can detect important in-game events, model athlete behavior, make movement predictions and more. It all helps dissect the minute details in sports, enabling teams to make smarter decisions on the field.
TVConal is a member of NVIDIA Inception, a free program that supports startups revolutionizing industries with cutting-edge technology.
Automated Match Tagging
Match tagging — creating a timeline of significant in-game events — is crucial to sports video analytics. Tags are used to generate detailed reports that provide performance statistics and visual feedback for referees, coaches, athletes and fans.
Since plays and other in-game events occur in mere instants, up to 20 loggers work together to accomplish live tagging for some sports matches, according to Izadi. This can be time consuming and labor intensive.
With TVConal’s platform, sports analysts can extract insights from video frames with just a few clicks — as AI helps to automatically and accurately tag matches in real time. This gives analysts the time to dig deeper into the data and provide more detailed feedback for teams.
The platform can also catch critical moments or foul plays that the naked eye might miss.
“If a player does an illegal action that’s beyond a human’s ability to process in a few milliseconds, the platform can detect that and inform the umpires to take an action just in time,” Izadi said.
TVConal’s platform is built using NVIDIA Metropolis, which simplifies the development, deployment and scale of AI-enabled video analytics applications from edge to cloud. Metropolis includes pretrained models, training and optimization tools, software development kits, CUDA-X libraries and more — all optimized to run on NVIDIA-Certified Systems based on the NVIDIA EGX enterprise platform for accelerated computing.
“NVIDIA’s software tools, frameworks and hardware allow us to iterate faster and bring ideas to market with shortened life cycles and reduced costs,” Izadi said.
NVIDIA GPU-accelerated compute resources used in TVConal’s platform include the NVIDIA Jetson platform for AI at the edge, RTX 3090 workstations on-prem and Tesla V100 and A100 in the cloud.
TVConal uses the NVIDIA DeepStream SDK to simplify video processing pipelines; NVIDIA pretrained models and the TAO toolkit to accelerate AI training; and the NVIDIA TensorRT SDK to optimize inference.
DeepStream enabled the TVConal team to process live video and audio streams in real time — 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.
And as a member of NVIDIA Inception, TVConal has access to technical resources, industry experts and go-to-market support.
The company’s clients include international production company NEP Group, the Pakistan Cricket Board and others.
“There is an increasing volume of sports content to extract value from,” said Izadi, highlighting that the global sports analytics market size is expected to grow over 20% by 2028. “Automated video processing is revolutionary in sports, and we are excited to build more advanced models and pipelines to keep the revolution going.”
Watch an on-demand NVIDIA GTC session about how AI is revolutionizing the sports industry — better predicting competition outcomes, improving performance and increasing viewers’ quality expectations.