Archive translation. This developer note was published in September 2021. APIs, model names, browser support and hosted demos may have changed; verify current TensorFlow.js and MediaPipe documentation before implementation.
In brief: TensorFlow introduced a free programming interface for real-time 3D pose detection in web applications.
Potential uses included browser-based health and fitness tools, medical research interfaces and motion-capture experiments. The published workflow combined TensorFlow.js with MediaPipe BlazePose and the GHUM body model.
Because processing could run in a web application, developers could prototype pose-aware experiences without a traditional desktop installation. Practical performance depended on browser support, camera input, hardware acceleration and the selected model configuration.
More information: The original technical explanation appeared on the official TensorFlow blog. A hosted demonstration was also provided as a web app; verify its current availability and privacy implications before granting camera access.
