Human Pose Analysis

AI capabilities are getting so powerful that we can now quite accurately determine and track a human skeleton overlay of the human form. This capability, often referred to as pose estimation, is surprisingly powerful as a starting point for analyzing the human form in imagery for myriad applications. For example, here at Xyonix we are helping a customer build models to facilitate early onset autism and cerebral palsy diagnoses in infants. Deep learning techniques using the infant pose estimation significantly outperform results applied to video frames themselves. In addition to assisting physicians with diagnoses, pose estimation can be used in sports and rehabilitation therapy applications like automatically analyzing posture, gait and other body poses. Human pose analysis can also be quite helpful in identifying areas of improvement for athletes, such as optimizing a runners form, a golfers swing, or a soccer goalies dives. Other key areas where pose estimation can be quite helpful include assisted living (fall detection), character animation (alleviates the need for costly sensors), video games, and animal behavior understanding.

Interested in applying human pose analysis to your problem? >> CONTACT US << and let’s dig in together. Want to find out more first? You can

read a case study on using AI and video to improve running performance through identifying body form related issues.

Curious how well the models perform but don’t want to dig into a deep analysis? Take a look at this video we made of shuffle dancers with superimposed skeletal forms. Advice: crank up the volume on your headphones first.