Across the world, sports are one of the strongest threads that bring people together. Over the years, innovations in technology have shifted how we play and experience games. Viewers at home increasingly look for new ways to understand how exactly a spectacular feat was accomplished just moments after they witness it.
After years of development, my team at wrnch has developed real-time 3D tracking using a computer vision and deep learning solution that can be deployed with any standard camera. This allows tracking of the motion and poses of athletes without them needing to wear any special suits and sensors.
Additionally with our solution utilizing 23 skeletal tracking points, a number of advanced metrics, such as velocity, stride length, and speed, can be reported live. Not only does this revolutionize the viewing experience on television, but also allows for athletes and coaches to gain access to performance-altering data like never before.
As an Intel® AI Builders member, we tested our 3D tracking solution, which we earlier thought was feasible only with GPUs, on Intel-powered CPUs and were blown away with the results. Using Future Generation Intel® Xeon® Scalable processors, codenamed Cascade Lake, to take advantage of the new Intel® DL Boost capability and Intel® Distribution of OpenVINO™ toolkit,we were able to improve inferencing speeds from 90 FPS to 450 FPS, a 5x increase*.
We’re thrilled with this improvement in performance and look forward to bringing this to market.
*System #1 (The wrnch standard hardware spec) – Intel® Core™ i7 processor, NVIDIA GTX 1080 gpu, 128 GB of DDR4-2666, inference resolution 328×184, Caffe* machine learning framework; System #2 – Future Generation Intel® Xeon® Scalable processor, codenamed Cascade Lake, 128 GB of DDR4-2666, Inference Resolution 328×184, Intel® Optimization for Caffe*, Intel® Distribution of OpenVINO™ toolkit
Software and workloads used in performance tests may have been optimized for performance only on Intel microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more information go to http://www.intel.com/performance/datacenter.