Instruction Platform Taps AI to Improve At-Home Yoga Experience

MixPose created an online yoga platform that utilizes AI and live video to make at-home workouts more connected. Both users and instructors can view pose estimations, and instructors can then give users helpful tips to improve their posture and overall yoga experience.

Image credit: MixPose

Many athletic activities offer online platforms, but the feedback loop between instructors and users was missing amid the pandemic. Peter Ma created MixPose, an interactive AI yoga platform, to better connect yoga instructors and users, as at-home exercise grows in popularity. MixPose uses AI for pose recognition, allowing users to see how their positions match up to the standard and allows instructors to give tailored feedback to users.

MixPose and Agora partnered to layer AI with video. MixPose users can join either a private, public, or semi-private channel, run by Agora’s interactive live video SDK. A private channel allows users to hide their video from other users, so only the instructor can view their video and give feedback. A semi-private channel allows users to share their video with select friends and the instructor, and a public channel shares the video with everyone on the feed.

Users can also hide their video entirely. This way, users of various skill and comfort levels can participate in MixPose’s yoga lessons. MixPose Chat, a feature on the MixPose platform, allows users to chat with others and the instructor.

Smart Technology Keeps Yoga Practices Flowing

The MixPose platform uses Intel® Deep Link technology, the Intel® Distribution of OpenVINO™ toolkit, and the x264 video codec application. The Intel® Deep Link technology prioritizes performance by routing power to both graphics processing units (GPUs) and central processing units (CPUs) for quick video encoding. The Intel® Distribution of OpenVINO™ toolkit, an AI inference system, also helps run GPUs and CPUs for better performance and development. With the use of the Intel® Iris® Xe MAX GPU, the existing CPU can focus on the MixPose operating system, improving efficiency for both processing units. The processing units work together to create an integrated online yoga experience.

Image credit: MixPose

Posture Detection and Video Encoding

MixPose uses the Intel® Iris® Xe MAX GPU, which allows MixPose to offer pose detection and efficient video encoding during yoga lessons. The GPU uses inference and image recognition to detect and analyze users’ poses and then classify them using AI. Lines and data points appear on users’ screens to show them their alignment.

The information is shown to instructors, in real time, so they can comment on poses and help fix any errors. The Intel® Iris® Xe MAX GPU works well on small computers or mobile devices, so it’s a great choice for at home yoga lessons.

MixPose users purchase a subscription for about $20 per month to attend classes and receive feedback, and users can choose from several different yoga instructors and yoga sessions. MixPose also offers certain free yoga classes as well. MixPose is available as a mobile app and also on a Web browser. Currently, MixPose only offers yoga classes, but the company plans to expand its technology to dance and personal fitness in the future as demand for connected at-home exercise increases.