Pose tracking

Real-time computer vision technology that tracks a subject's 2D/3D body position by identifying key anatomical landmarks (joints) across video frames.

Pose Tracking, or Human Pose Estimation (HPE), uses deep learning models like MediaPipe Pose (BlazePose) or YOLO11 Pose to map a person's movement: the system identifies a set of keypoints (e.g., 17 to 33 landmarks) on the body, including hands, feet, and face, creating a skeletal representation. This technology delivers real-time performance, with models like YOLO11 Pose achieving 89.4% [email protected] on COCO Keypoints and running at 30+ FPS on mobile devices. Key applications are diverse: real-time fitness coaching (form correction), sports analytics (biomechanical analysis), and AR/VR motion capture.

https://developers.google.com/ml-kit/vision/pose-detection