Putting the capabilities of Intel® Deep Learning Boost (Intel® DL Boost) into more hands (and arms, legs, torsos).
Figure 1. Woman doing a plank with visual overlay.
Video recording technology is nothing new, but it has come a long way in the last few decades, moving from the hands of very few to a tool for the masses. More recently, motion capture technology has been put to use at the highest levels in Hollywood and in professional sports, improving audience experiences and athletic performance, respectively. Until now, such technology has been out of reach for most consumers, requiring expensive, specialized equipment. wrnch* is challenging that with AI driven, marker-less motion capture technology that can be used by any 2D webcam.
“Using Intel® Deep Learning Boost (Intel® DL Boost) through the OpenVINO™ toolkit, our developers worked with Intel engineering to optimize wrnch to run great on 10th generation Intel® Core™ i5 processors.”
— Paul Kruszewski, CEO, wrnch
wrnch can turn any camera into, essentially, a visual cortex, that translates visual information. By helping computers to make connections between what they see and what they know, we are teaching computers how to better understand human motion, activity, and body language. From entertainment to security, the application potential for this is widespread. The question of where to take the technology falls to the creative minds of today’s independent software vendors.
By combining motion capture technology with computer vision, coaches and trainers can access powerful insights at a level of precision unavailable to human senses. This wealth of information allows for more precise training, which ultimately results in better athletes. Unfortunately, this technology has been expensive and inaccessible to the rest of the public before now. But amateur athletes come in all shapes and sizes. By removing the need for wearable sensors and allowing the computer to see more naturally, and using only a standard webcam, computer vision and training can be harnessed in the home gym.
Figure 2. Woman doing a lunge with visual overlay.
Exercise is a multimillion dollar industry that crosses every demographic, thanks in part to sports and rehabilitation. As a use case demo for this developing technology, wrnch has been testing a virtual personal trainer application for the software. Personal trainers can be economically unrealistic for a lot of the population. Workout videos have long been a popular alternative for those who prefer to do lunges in the privacy of their own rooms. More recently, thanks to virtual assistants, workout apps are making regular exercise even easier to work into a daily routine. In the future, wrnch would like to see workout apps that use computer vision to count reps and to help people ensure that they are doing exercises correctly. It is one thing to watch a workout host pull a perfect plank, but quite another to have your home computer capable of telling you when you need to straighten your legs.
Exercise, and exercising correctly, goes beyond physical fitness. The same personal training program that can talk a yoga practitioner into an improved tree pose could be used for home rehabilitation exercises after a hip or knee replacement. Physical therapists would tell you that exercises work best when performed in a particular way. Using the integrated webcam on an Intel® Core™ i5 processor powered laptop, wrnch can be applied to help ensure a body is in proper alignment for maximum safety and results. In a similar vein, a computer can be trained to recognize the physical signs of a person who is in distress and needs assistance. This application could potentially be helpful in homes of the elderly and other demographics at risk of home injury.
The far-reaching applications of wrnch will be in the hands of today’s talented developers, whose creativity will doubtlessly find new uses for the real-time, marker-less motion capture technology. In the developing field of robotics, this virtual visual cortex is already helping computers learn to interpret human behavior in order to create more natural and helpful interactions. Even simple human cues, like body positioning, can help a robot to understand whether a human is trying to interact with it. Similarly, the actions and body language of pedestrians could help self-driving cars recognize when a person might try to cross the street.
From a security standpoint, a computer’s ability to quickly spot and interpret suspicious behavior could go far to keeping public and private spaces more secure, from banks, to airports, and schools. As robots take on more human-adjacent roles, their ability to understand and predict our actions will become increasingly critical to successful adoption.
wrnch is powered by Intel® processors using Intel DL Boost through the OpenVINO toolkit, which is on 10th generation Intel® Core™ processors. For more information on Intel DL Boost and AI-powered apps see wrnch and OpenVINO toolkit.
Intel's compilers may or may not optimize to the same degree for non-Intel microprocessors for optimizations that are not unique to Intel microprocessors. These optimizations include SSE2, SSE3, and SSSE3 instruction sets and other optimizations. Intel does not guarantee the availability, functionality, or effectiveness of any optimization on microprocessors not manufactured by Intel. Microprocessor-dependent optimizations in this product are intended for use with Intel microprocessors. Certain optimizations not specific to Intel microarchitecture are reserved for Intel microprocessors. Please refer to the applicable product User and Reference Guides for more information regarding the specific instruction sets covered by this notice.
Notice revision #20110804