Live data analysis for spots is a tough sector to tackle. Not only there are movements all around the entire playfield but crunching data on the fly with anything possible occurring in the next second is difficult for even the best computing hardware. But it is about to change as Singapore-based organization TVConal utilizes some industrial magic provided by the NVIDIA Metropolis platform.
Although Team Green’s video analytics platform is initially engaged to help drive a smart city vision by delivering insights for the sake of improving aspects such as traffic and public safety – catching close to the ctOS city management system from the Watch Dogs video game, the roots of the AI algorithm is still vision-based and with some smart application, it can be turned and utilized for other purposes as well. As the processing platform is capable of analyzing fast-moving objects, sports is no doubt one of the most useful sectors to deploy the tech.
Rolling back to the topic of TVConal, the company name actually stands for Television Content Analytics and it has been a member of NVIDIA’s Inception program for quite some time where the GPU giant supports starts with bleeding-edge technology revolving around AI and computer vision. According to their announcement, the NVIDIA Metropolis integration will be applied to some of the more popular sports in Asia including but not limited to badminton, cricket, football, and tennis, with possible additions down the line.
Taking a page out of the examples section, AI vision is capable of detecting important moments in a sports event that otherwise could be missed by our human eyes due to distractions such as illegal moves going unnoticed or player pattern dissection through behavior predictions and tactics analysis – all of which can be captured and dissected by a computer frame by frame. Not only that, a lot of the tasks can be automated as well such as match tagging, a slew of work that records and labels important events of a game, so that broadcasters and hosts can focus on casting the game full on.
Details about the newly added NVIDIA GPU-accelerated computing resources show a combination of various dedicated hardware such as NVIDIA Jetson for AI edge computing, GeForce RTX 3090 equipped workstations for on-premise processing while the Tesla cards of the V100 and A100 are in charge of cloud computing. Software is also Team Green’s specialty when it comes to optimizations as the DeepStreak SDK simplifies video processing pipelines while pre-trained models and the TAO toolkit can cut down the amount of time required for specifically supervised learning. The TensorRT SDK and library are also a great help too as the TVConal team can process video and audio streams in real-time fast enough while maintaining high accuracy.
As the power of machine learning gets explored furthermore and seeps deeper into large-scale commercial deployment, it seems that a true smart city isn’t really that far from realization.